85. AI Employees
In this episode we discuss: AI employees. We are joined by Matt Lhoumeau, Cofounder & CEO at Concord
Love The Operations Room? Please support us by rating and reviewing it here.
We chat about the following with Matt Lhoumeau:
- How do you balance the need for structure with the chaos of fast-growing operations?
- What happens when marketing becomes a true conversation rather than a one-way message?
- Why are standards and expectations so critical to building trust in a growing team?
- How can leaders create clarity without stifling creativity?
- What’s the real difference between being busy and being effective?
References
- https://www.linkedin.com/in/mattlhoumeau/
- https://www.concord.app/
- https://eu1.hubs.ly/H0lMWln0
Biography
Matt Lhoumeau is the CEO and co-founder of Concord, the leading provider of AI-powered Agreement Intelligence solutions. With over a decade of experience transforming how businesses manage contracts, Matt helps operations leaders unlock strategic value from their agreements and turn contracts from cost centers into profit drivers.
To learn more about Beth and Brandon or to find out about sponsorship opportunities click here.
Summary
04:05 – AI tools and operational leadership.
08:05 – The power of systems thinking in AI bots
12:01 – Using AI to challenge you
15:16 – AI-first mantra
19:40 – Legal department AI substitutes
22:36 – Being a systems owner
24:30 – ChatGPT as a therapist
27:37 – Using AI to evolve and be durable
31:34 – How tools and frameworks are used over time.
34:58 – The balance between AI and human intervention
35:19 – Evolving with AI
37:05 – The right inputs
This podcast uses the following third-party services for analysis:
Podcorn - https://podcorn.com/privacy
Transcript
Hello everyone and welcome to
Speaker:another episode of the operations
Speaker:room a podcast for COOs.
Speaker:I'm Brandon Mensinga joined by my
Speaker:amazing co-host Bethany Ayers.
Speaker:How are things going Bethany this
Speaker:morning?
Speaker:I'm well.
Speaker:I had our first
Speaker:company off site this week,
Speaker:Tuesday and Wednesday.
Speaker:It was really good.
Speaker:We randomly had a voucher
Speaker:from a canceled or
Speaker:a postponed off site from last year,
Speaker:where the team were going to go to
Speaker:Serbia for
Speaker:four days, three nights.
Speaker:Ended up having to postpone it
Speaker:because a large customer win came
Speaker:in and like too much of the
Speaker:team needed to be involved in that
Speaker:to go away for four days.
Speaker:And so we had this voucher to spend
Speaker:and they really wanted us to go back
Speaker:to Serbia. Like I think they
Speaker:might've owed the Serbian
Speaker:place some money.
Speaker:And there was just no way it was
Speaker:going to take four days out of the
Speaker:business. And also it was
Speaker:getting to Belgrade and then a two
Speaker:hour transfer to the middle of the
Speaker:mountains.
Speaker:What? Who decided to do this?
Speaker:So A, why Serbia?
Speaker:B, why two hours out of the airport?
Speaker:And you have a voucher as a result.
Speaker:Yeah, so it was like
Speaker:seven hours of transport.
Speaker:So you're basically spending two
Speaker:days traveling in order to spend two
Speaker:days together.
Speaker:So we switched
Speaker:it to Brighton and
Speaker:one night in Brighton cost the
Speaker:same as three nights in Serbia.
Speaker:I can imagine.
Speaker:But also like transport
Speaker:was way cheaper.
Speaker:So overall NetNet we were definitely
Speaker:winners going to Brighton.
Speaker:Nice hotel, really easy
Speaker:for everyone to get to.
Speaker:We had quite the night out
Speaker:or the team had quite a night out on
Speaker:Wednesday night.
Speaker:Like I purposely went to bed at half
Speaker:10. One, I'm just not much of a
Speaker:drinker anymore.
Speaker:And two, everybody can have more
Speaker:fun when the boss isn't around
Speaker:and like.
Speaker:Three CEOs in six months
Speaker:is a lot for the team.
Speaker:There's a lot of change.
Speaker:Just let everybody blow off a bit of
Speaker:steam. But the blowing off a big of
Speaker:steam meant that some members
Speaker:of the team did not go
Speaker:to bed till 4.30 in the morning.
Speaker:What?
Speaker:And you have the off-site the next
Speaker:day, and they're going to bed at
Speaker:4.30.
Speaker:Wow. Like the second day of the
Speaker:offsite, yeah.
Speaker:The second day. Okay, fine.
Speaker:And the hotel was right
Speaker:on the beach.
Speaker:If you've been to Brighton, there's
Speaker:like a big road and then we were
Speaker:the very first hotel there and
Speaker:members of the team just had to go
Speaker:and get fresh air on a fairly
Speaker:regular basis. They were looking
Speaker:very peaky.
Speaker:They'd go get some fresh sea air for
Speaker:20 minutes, come back looking
Speaker:fine. And about half an hour later,
Speaker:they were just collapsing again in
Speaker:their peakiness.
Speaker:I was just entertained.
Speaker:So the bulk of us were able to
Speaker:build our AI employees reason.
Speaker:Test a bunch of technology and
Speaker:have a bit of a, not a hackathon
Speaker:because everything that was built
Speaker:needs to be able to be used.
Speaker:But then there was like a handful of
Speaker:people who used the second
Speaker:day to breathe in sea air.
Speaker:Okay, so pull us back for a second.
Speaker:So two day off site, you're building
Speaker:AI agents.
Speaker:What is going on here?
Speaker:So, one of our OKRs
Speaker:for the quarter is to become an AI
Speaker:first company, and our
Speaker:key results for that is
Speaker:to have 50 AI
Speaker:employees in the business
Speaker:and then have
Speaker:end-to-end automation
Speaker:in the go-to market team and an end-
Speaker:to-end animation in the engineering
Speaker:team. Go-to market have already done
Speaker:it, so we might have to make that
Speaker:key result a bit harder.
Speaker:And then on the engineering
Speaker:team, they have an idea.
Speaker:Which is basically linear
Speaker:will talk to Claude
Speaker:code and
Speaker:fix small bugs automated.
Speaker:So you log a bug
Speaker:and then Claude Code goes off and
Speaker:fixes it.
Speaker:So they're gonna test that and see
Speaker:for a lot of like just the small,
Speaker:like simple ones.
Speaker:And then on the go-to-market team,
Speaker:we have built, I don't
Speaker:know like the timings of
Speaker:these episodes that what Donna
Speaker:McCurley talked about in terms of.
Speaker:Figuring out, we have one agent
Speaker:that looks at what our best
Speaker:customers look like and goes off and
Speaker:finds look-alikes with the buying
Speaker:intent that then
Speaker:writes personalized emails
Speaker:based on what good looks like
Speaker:and then enters them into
Speaker:a HubSpot sequence.
Speaker:And then from responses to the
Speaker:sequence, it feeds the SDR.
Speaker:That is super awesome.
Speaker:And then what are you using for the
Speaker:tooling to make that happen?
Speaker:So we now have business
Speaker:chat GPT for everyone in the team,
Speaker:and we are
Speaker:using in go-to-market a tool
Speaker:called Cargo, which is kind
Speaker:of like an N8N or a
Speaker:Clay, but specific, like it's
Speaker:like Clay in that it's specific to
Speaker:go-go-to market.
Speaker:But from what I understand,
Speaker:it can also do N8n or
Speaker:Zapier type stuff.
Speaker:So even though it happens to have
Speaker:some nice plugins for
Speaker:like dealing with LinkedIn.
Speaker:Finding lookalikes, different areas
Speaker:that it can search over.
Speaker:It also then just does automated
Speaker:workflows like any of these tools.
Speaker:And then we also bought
Speaker:Claude code subscriptions
Speaker:for everybody in the
Speaker:engineering team.
Speaker:We had cursor and the
Speaker:cursor you can put different things
Speaker:in, but apparently Claude Code is
Speaker:just like, it's a whole new thing
Speaker:and it's way better.
Speaker:And it's not just the model that you
Speaker:plug into Cursor.
Speaker:What Cloud Code can do is
Speaker:astounding.
Speaker:And we had training on
Speaker:day one, because I'm
Speaker:bouncing over a place, so I can tell
Speaker:I'm a bit tired.
Speaker:So we are working on
Speaker:creating these AI employees and
Speaker:becoming the AI first company.
Speaker:And so we use the offsite to ring
Speaker:fence time for everybody to
Speaker:learn and explore.
Speaker:And so day one before
Speaker:the fun.
Speaker:Letting off steam till 430
Speaker:in the morning evening.
Speaker:Everybody was sober and not
Speaker:hung over and capable of being in
Speaker:training for the day.
Speaker:We did two
Speaker:training sessions.
Speaker:So the morning session was for the
Speaker:entire company.
Speaker:What do you call it?
Speaker:Not when you're the number one
Speaker:fan of somebody like My idol?
Speaker:You're a fanboy, a fangirl in this
Speaker:case.
Speaker:I'm a fangirl, yeah, of
Speaker:Charlie Cowan, who's been on the
Speaker:podcast a couple times.
Speaker:Charlie ran his
Speaker:101 and 102 classes
Speaker:on ChatGPT for the whole
Speaker:team.
Speaker:The most amazing feedback.
Speaker:Everybody in the team loved it,
Speaker:including the engineers, because I
Speaker:thought it might just be a bit basic
Speaker:for the engineers.
Speaker:But everybody learned, everybody
Speaker:loved Charlie's energy.
Speaker:And at the wrap up
Speaker:on the second day.
Speaker:Multiple people asked if we could do
Speaker:quarterly training with him.
Speaker:Oh, wow.
Speaker:OK. Kudos, Charlie.
Speaker:That's amazing.
Speaker:I have never had training
Speaker:in a company where people
Speaker:were begging for more training.
Speaker:Yeah, that's super awesome.
Speaker:And we all just learned so much
Speaker:around what chat GPT can do,
Speaker:all the different functionality,
Speaker:ways to approach it.
Speaker:He had a very safe environment.
Speaker:Lots of people would own up to
Speaker:mistakes they made or problems they
Speaker:were having.
Speaker:He gave different techniques.
Speaker:And then in the afternoon,
Speaker:the go-to-market team went and
Speaker:started building their
Speaker:AI employees had a brainstorm on
Speaker:like. What kinds of things
Speaker:could be built, what would be
Speaker:helpful in their jobs, how to build
Speaker:them, and started doing that.
Speaker:And the afternoon was a combination
Speaker:of Charlie and somebody else who's
Speaker:going to be a guest on our show
Speaker:soon, Ryan Fuller.
Speaker:And we did a more like workshoppy
Speaker:thing and more
Speaker:of a deep dive into Cloud Code.
Speaker:But it wasn't as structured
Speaker:a format as Charlie did
Speaker:for Chat GPT.
Speaker:So I'm thinking for the next
Speaker:quarterly training, we do a deep
Speaker:on cloud code, because apparently...
Speaker:It can do stuff even for non-coders.
Speaker:Like Charlie's using it to run his
Speaker:strategy.
Speaker:Charlie used it to help him
Speaker:build his new website.
Speaker:Like the way that the structure is,
Speaker:I don't entirely understand, means
Speaker:that it's super powerful, not
Speaker:just for coding.
Speaker:That's amazing.
Speaker:So then, Charlie Cowan, 101,
Speaker:102, the sessions up front for
Speaker:ChatGBT, this fellow doing more
Speaker:of a workshop on cloud code in
Speaker:day one, and then you're doing a bit
Speaker:of like brainstorming with the
Speaker:functional teams around AI employees
Speaker:and having the teams think through
Speaker:how to build those things as part of
Speaker:the day one.
Speaker:As part of day one, and then day two
Speaker:was building things.
Speaker:And then we all got
Speaker:together, and there was a bit of a
Speaker:show and tell for anybody who wanted
Speaker:to show what they'd built and
Speaker:how they'd build it.
Speaker:So I built a
Speaker:demo planner, GPT,
Speaker:where you can drop in transcripts
Speaker:from your first meet,
Speaker:and it will help you build a
Speaker:tailored demo.
Speaker:I put in the Sandler methodology,
Speaker:MedPic.
Speaker:Another methodology on like what
Speaker:good demos look like,
Speaker:some of our transcripts anonymized
Speaker:around first meets and
Speaker:good demos so it knows what good
Speaker:looks like, our knowledge
Speaker:base so it understands our platform
Speaker:and how it works, and I can't
Speaker:remember a couple other things, plus
Speaker:the instructions.
Speaker:ChatGPT helped me write the
Speaker:instructions for the GPT
Speaker:on the structure of what it should
Speaker:look like the questions you should
Speaker:ask, what the outputs are, and
Speaker:then When you drop in
Speaker:a transcript, it tells you
Speaker:a summary of the calls of who the
Speaker:key decision makers are, where the
Speaker:power is, all of the stuff that like
Speaker:between Sandler and MedPic you get.
Speaker:It gives you the flow for the
Speaker:demo, questions to drop
Speaker:in and pepper, talking points
Speaker:and which person to address because
Speaker:that was their particular need in
Speaker:the demo. A follow-up
Speaker:email, and a
Speaker:mini- mutual action
Speaker:plan to get to the proof of value
Speaker:because it also understands our
Speaker:sales process.
Speaker:So it's like, so we need to do a
Speaker:proof of value in three weeks.
Speaker:Therefore, this step needs to
Speaker:happen. These are the owners and
Speaker:it's ready to go that you can just
Speaker:share with a customer.
Speaker:That's amazing.
Speaker:And then is this a folder in
Speaker:ChachiBT that you're using to do
Speaker:this or?
Speaker:So this is a GPT
Speaker:that I've used on this one.
Speaker:So you, a custom GPT,
Speaker:that's just shared within Matomic.
Speaker:Cause now that we have the
Speaker:GPT workspace, you can
Speaker:publish things and everything can be
Speaker:available to everybody else.
Speaker:So can talk into the GP
Speaker:T and it has all of this structure
Speaker:and then you get your own chat.
Speaker:And then that chat will have your
Speaker:memory in it of just your
Speaker:chat. Whereas the GPt doesn't get
Speaker:affected by other people's memory.
Speaker:If that makes sense.
Speaker:That makes sense.
Speaker:Shareable custom GBT.
Speaker:You can do your own work in there
Speaker:and maintains a memory persistence
Speaker:thing whereby it understands your
Speaker:context specifically, but doesn't
Speaker:pollute others with yours.
Speaker:Exactly, because it's in like a
Speaker:separate chat with yours, but the
Speaker:GPT itself has not been polluted by
Speaker:your content.
Speaker:And it can also, if you're doing a
Speaker:first meet, you can give a little
Speaker:bit of background and ask for
Speaker:questions to make sure to
Speaker:ask so that it'll give you a good
Speaker:tailored demo for the second meet.
Speaker:So that's what I built.
Speaker:And then I'm in the process
Speaker:of building a CMO
Speaker:for us using some
Speaker:content of somebody online that
Speaker:Charlie recommended.
Speaker:Our CSM team.
Speaker:Connected all of their different
Speaker:like HubSpot plus
Speaker:other spreadsheets they have of all
Speaker:kinds of data.
Speaker:And they can now query when
Speaker:a renewal is coming up, what's our
Speaker:forecast, what are my
Speaker:actions, which QBRs do I
Speaker:need to do?
Speaker:And also it goes and looks
Speaker:at the news for
Speaker:your customers.
Speaker:So it could give you actions plus
Speaker:news that you need to know about
Speaker:like in the outside world and little
Speaker:snippets to share with them.
Speaker:In retrospect, what's your sense
Speaker:of the value of the two-day
Speaker:session itself?
Speaker:What do you have done anything differently?
Speaker:Do you think this is really the
Speaker:springboard to become, to achieve
Speaker:your OKRs and be AI first?
Speaker:Oh, totally. We needed those
Speaker:days of time outside
Speaker:of your day-to-day work to just
Speaker:open people's minds and the
Speaker:training.
Speaker:Because actually, one of the things
Speaker:that I learned that I wasn't
Speaker:expecting is I tried the chat
Speaker:GPT voice
Speaker:mode before and it was quite
Speaker:slow and didn't feel like a human.
Speaker:It's massively improved.
Speaker:So Charlie demoed it and then
Speaker:yesterday in one-to-one
Speaker:with an SDR.
Speaker:So we've created another
Speaker:custom GPT that is
Speaker:a cynical CISO.
Speaker:So we'd done like a persona of a
Speaker:chief information officer and we run
Speaker:all of our content through that to
Speaker:make sure that we are not too
Speaker:salesy, that we seem credible.
Speaker:And we just flipped that into chat
Speaker:mode, said,
Speaker:we're calling you, here's the
Speaker:first pitch.
Speaker:How can we make it better?
Speaker:And because ChatGPT is always very
Speaker:sensible, it's like, oh, well that
Speaker:was a really good pitch.
Speaker:These are, you know, some notes I'd
Speaker:have on how to improve it.
Speaker:And it came up with like, it was an
Speaker:excellent pitch and it feels
Speaker:like an actual person.
Speaker:So you can really practice your
Speaker:pitches. You can practice going out
Speaker:to investors.
Speaker:You can practice everything.
Speaker:Charlie and I were chatting this
Speaker:morning and he'd had another
Speaker:training session with another team
Speaker:of execs yesterday and
Speaker:they're going out for fundraising
Speaker:and the CEO, they built
Speaker:a grumpy investor.
Speaker:And we're just practicing pitching
Speaker:to a grumpy investor like what
Speaker:landed what didn't land what would
Speaker:be better ideas amazing
Speaker:for anybody who's doing human
Speaker:stuff like how to practice without
Speaker:having to be on other humans.
Speaker:I think one of the things that I'm
Speaker:thinking about right now is this
Speaker:question of time and space to like
Speaker:think about the stuff.
Speaker:Because I feel like in my current
Speaker:company and previous companies,
Speaker:the day-to-day grind of getting
Speaker:stuff done and what has to get done
Speaker:for the business doesn't allow you
Speaker:to have a lot of extra space and
Speaker:capacity to think about different
Speaker:stuff, you know, and in particular
Speaker:for AI to your point, it takes
Speaker:a bit of time in effort to think
Speaker:through like if we wanted to do
Speaker:something here properly with AI
Speaker:employees or what have you or for
Speaker:myself or for the team, it requires
Speaker:space to do that.
Speaker:So this idea of doing a
Speaker:proper off-site in the format that
Speaker:you just talked about Feels like
Speaker:that's almost like an instrumental
Speaker:linchpin to like get
Speaker:everyone's mind in the same place,
Speaker:get the training there to enable
Speaker:them to be successful creating these
Speaker:employees and set yourself on
Speaker:an actual path.
Speaker:100% because also like
Speaker:so many times companies waste days
Speaker:on hackathons and like it's
Speaker:quite nice to have a hackathon
Speaker:because you can experiment but
Speaker:it all ends up just being
Speaker:shelf wear and so
Speaker:it was like this was not
Speaker:around experimenting I mean
Speaker:it was because you're always going
Speaker:to get better but it needed to be
Speaker:things that would actually make
Speaker:your lives better tomorrow and
Speaker:figure stuff out and also like
Speaker:playing around with AI employees,
Speaker:but also playing around with or
Speaker:taking the time to get
Speaker:our HubSpot data better, get
Speaker:our dev environment faster to move.
Speaker:Like there's almost ring-fencing
Speaker:time to slow down with
Speaker:all of those things that you know if
Speaker:you do, you'll speed up, but aren't
Speaker:quite bad enough.
Speaker:So I think we're going to do this at
Speaker:least quarterly.
Speaker:It might not be an offsite every
Speaker:time, but ring-fence
Speaker:time to slowly down to
Speaker:speed up.
Speaker:Specifically around AI and
Speaker:automation.
Speaker:And then the overall mindset
Speaker:and motivation, morale,
Speaker:enthusiasm of the company now
Speaker:with this AI first mantra that
Speaker:you've brought to the company, can
Speaker:you sense a real momentum now off
Speaker:the back of this in the sense that
Speaker:folks are really enthused and
Speaker:excited to move forward with it?
Speaker:Yes.
Speaker:My hesitation is just that there's
Speaker:always a spectrum.
Speaker:You have some people who have a
Speaker:hundred percent embraced it.
Speaker:We have one engineer
Speaker:in particular who has just opened
Speaker:his mind and he is doing
Speaker:so much.
Speaker:And you have others who are still a
Speaker:little bit skeptical.
Speaker:They're into it.
Speaker:They're not defensive,
Speaker:but they're not as
Speaker:just totally excited.
Speaker:And then on the go-to-market side, I
Speaker:would say it's similar.
Speaker:For some people, there's just a
Speaker:click. And they're just like,
Speaker:this is amazing. And for others,
Speaker:that click hasn't 100% happened yet,
Speaker:but they're curious,
Speaker:want to see how it goes,
Speaker:and just need a little bit more
Speaker:support to have that aha moment.
Speaker:Here's a thought that I was having
Speaker:the other day.
Speaker:So there's another CEO, a
Speaker:friend of mine, that works in a
Speaker:organization, they're currently
Speaker:doing fundraising, and they've
Speaker:put together their financial
Speaker:forecast of their shopping route
Speaker:with investors in terms of here's
Speaker:our revenue, the assumptions, here's
Speaker:your budget and so on.
Speaker:So that individual has effectively
Speaker:shaved.
Speaker:Outside of the classic B2B SaaS
Speaker:scaling for headcount that
Speaker:we had previously, they've shaved
Speaker:off, let's say, 30%, 40%
Speaker:of the headcount.
Speaker:They've taken that cash and
Speaker:mostly put it into a line item
Speaker:called IT software.
Speaker:So per person, per GTM
Speaker:person, for developer allocating
Speaker:a spend amount annually for
Speaker:individuals in the company, and
Speaker:the amounts that are being put into
Speaker:the line item are quite significant.
Speaker:We don't know exactly how much is
Speaker:the right amount right now, but just
Speaker:from a forecasting standpoint, let's
Speaker:just take some of that headcount
Speaker:spend reallocated to this line item
Speaker:for the time being with really not
Speaker:knowing how the future is going to
Speaker:pan out, but let's not minimize the
Speaker:potential cost that might be
Speaker:associated with significant
Speaker:token spend as it were, in
Speaker:particular for developers as they
Speaker:start using tooling quite
Speaker:aggressively going forward.
Speaker:What do you think of that?
Speaker:I think it's wise.
Speaker:I think, it's also generous.
Speaker:Although it would future proof
Speaker:you because everything right now is
Speaker:a loss leader.
Speaker:Nobody's making money and prices are
Speaker:going to have to go up.
Speaker:And as people move into the
Speaker:enterprise, prices are gonna go up,
Speaker:I kind of hoping that the enterprise
Speaker:will subsidize the
Speaker:startups. I would say it's probably
Speaker:wise to forecast a bit more
Speaker:spend so that when the prices do
Speaker:ratchet up, you have the money?
Speaker:So as we have
Speaker:just bought Claude code for
Speaker:everybody, right now they have three
Speaker:subscriptions. They have a 20 pound
Speaker:one that year, a $20 one that you
Speaker:apparently run out of within
Speaker:seconds, a
Speaker:hundred dollar one that
Speaker:seems to work for most people and
Speaker:then a 200.
Speaker:And we've gone for the hundred and
Speaker:see if that works.
Speaker:Otherwise go to the 200.
Speaker:With the API calls, you don't know
Speaker:what's going to happen in the
Speaker:future. At least you know there'll
Speaker:be enough money.
Speaker:I think on the go-to-market side,
Speaker:it'll clear out a bit.
Speaker:Yeah, like right now, there's so
Speaker:many tools coming in, and I think
Speaker:quite a few of those tools will come
Speaker:back out again.
Speaker:And then I also think that
Speaker:chat GPT or OpenAI will
Speaker:end up building a lot of things that
Speaker:let you consolidate.
Speaker:Yes, I suspect so.
Speaker:If I were a note taker, I'd be very
Speaker:worried.
Speaker:So we have a topic for today, which
Speaker:we are talking about right now,
Speaker:which is AI employees.
Speaker:We have an amazing guest for this,
Speaker:which Matt Lumo.
Speaker:He is the CEO and co-founder of
Speaker:Concord.
Speaker:And we had a long conversation with
Speaker:Matt around contracting
Speaker:for companies and a bit of
Speaker:a chat around CS.
Speaker:But the interesting bit that we
Speaker:spoke about was about AI employees
Speaker:and what Matt is doing in that
Speaker:respect. So before we get into the
Speaker:AI employees piece of it,
Speaker:is there any thoughts just on his
Speaker:other bits that you talked about
Speaker:related to CS and
Speaker:contract management.
Speaker:I guess on the contract management
Speaker:part, I think I
Speaker:will, at some point,
Speaker:have a look at Concord, their
Speaker:technology, because how
Speaker:much do we actually need lawyers for
Speaker:for contracts versus how much can
Speaker:we use AI?
Speaker:And they've basically created a
Speaker:system where you can do a lot of
Speaker:your contract negotiation via
Speaker:their platform and then also not
Speaker:lose all of your contracts.
Speaker:I haven't tried it yet, but we
Speaker:don't have a lawyer in-house as we
Speaker:go up market and that are dealing
Speaker:with larger.
Speaker:Institutions, having somebody
Speaker:help us review contracts for
Speaker:standard ones seems like a good
Speaker:idea.
Speaker:Right now, for
Speaker:commercial contracts in my current
Speaker:company, we're just about to kick
Speaker:off a process to make this
Speaker:as friction-free and as fast as
Speaker:humanly possible, leveraging tools
Speaker:and a revised process.
Speaker:Right now we've been using a lot of
Speaker:human legal powers and work to
Speaker:make modifications to contracts, but
Speaker:it's all been very sloppy, I think,
Speaker:in the sense of just relying upon
Speaker:a person to inject themselves
Speaker:to make these changes and not really
Speaker:thinking through to make this
Speaker:a friction-free process that is
Speaker:much, much faster and doesn't lean
Speaker:on that person so much.
Speaker:The interesting bit is here
Speaker:in the UK, with some of these
Speaker:providers, there's some good
Speaker:providers out there. We've mentioned
Speaker:before like Harbor James and so on,
Speaker:where they're very SaaS friendly and
Speaker:they have subscription models
Speaker:whereby they're actually reasonably
Speaker:priced, I think, to be honest, like
Speaker:really good value.
Speaker:Talking to this woman in the US, I
Speaker:don't know if this is just like a US
Speaker:thing overall or just her
Speaker:specifically, but immediately after
Speaker:all this scale up talk, we went
Speaker:into what is your business
Speaker:model, how do you charge
Speaker:effectively?
Speaker:Oh, we charge by the hour.
Speaker:So it's very transactional, hourly
Speaker:based. And by the way, it's 750 US
Speaker:per hour.
Speaker:That is a lot of money.
Speaker:So let me get this straight.
Speaker:You're scale up focused.
Speaker:SaaS focused and you have this
Speaker:legacy old school business
Speaker:model charging hourly defeats
Speaker:the entire purpose because we want
Speaker:fractional embedded time
Speaker:consumption where you are
Speaker:representative as part of our
Speaker:company and we would like to pay you
Speaker:in a way that is not transactional.
Speaker:Sounds like you really need to talk
Speaker:to Matt, given that you're like
Speaker:streamlining everything and you're
Speaker:looking at using, you know, and
Speaker:like, how much of it can you use for
Speaker:AI?
Speaker:Well, so after this, I think you and
Speaker:Matt need to have a conversation,
Speaker:basically.
Speaker:So our conversation with Matt was
Speaker:for me wildly valuable, because
Speaker:I had never heard of the concept of
Speaker:AI employees until we chatted with
Speaker:him. And obviously, it has changed
Speaker:my worldview.
Speaker:And it has meant that these are
Speaker:the OKRs that we're having.
Speaker:And Concord are ahead of
Speaker:us on doing it, and the
Speaker:productivity that he's seeing out of
Speaker:his team is astounding.
Speaker:He's also a bit further ahead
Speaker:than us in that he doesn't have a
Speaker:CMO and a CRO anymore.
Speaker:He has go-to-market engineers.
Speaker:There's definitely a lot of hype
Speaker:around go- to-market-engineers.
Speaker:I've seen it on LinkedIn.
Speaker:It's the job of the future, and
Speaker:then for some people it's like it's
Speaker:already the job of the past and
Speaker:proven wrong. So I think the verdict
Speaker:is definitely still out for it.
Speaker:But what is the role
Speaker:of management?
Speaker:In the future when everybody
Speaker:has AI employees and
Speaker:live coaching is interesting.
Speaker:The SDR was getting way
Speaker:better pitch coaching out
Speaker:of the voice mode of our
Speaker:CISO GPT than from
Speaker:me or from a sales
Speaker:leader.
Speaker:I think one of the ideas that
Speaker:Matt had introduced was that
Speaker:the role of a manager, an actual
Speaker:human being in this case, being a
Speaker:manager becomes much more about
Speaker:being a systems owner and not
Speaker:being a people coach.
Speaker:That the people coach is actually
Speaker:left to the AI part of it, which is
Speaker:ironic in some ways.
Speaker:But as a systems' owner, your job is
Speaker:to create AI
Speaker:employees to onboard them,
Speaker:contextualize them, train them.
Speaker:Iterate them to make them better,
Speaker:give them performance reviews, give
Speaker:them kind of OKRs and results
Speaker:that need to be achieved and so on.
Speaker:Taking the time and effort to not
Speaker:just do a one-off kind of GPT
Speaker:prompt, but in fact care-take
Speaker:a given GPT
Speaker:employee to get the most out of
Speaker:them. That whole mindset shift
Speaker:is kind of what he'd spoken
Speaker:about, in particular for creating
Speaker:coaches. I can see that, yeah,
Speaker:being quite effective.
Speaker:Yeah, there's two points
Speaker:to make. One is,
Speaker:I think we need to be careful on
Speaker:what we talk about
Speaker:for like coaching versus
Speaker:in a school context, they call it
Speaker:pastoral care.
Speaker:So like the check-ins that you're
Speaker:doing okay, and maybe
Speaker:not coaching on figuring out a
Speaker:problem at work, but like the
Speaker:human contact.
Speaker:And I think that's still going to be
Speaker:important.
Speaker:And it shouldn't be
Speaker:outsourced to an
Speaker:AI employee.
Speaker:You know, as a CEO, I'm worried
Speaker:about what's the morale
Speaker:like? How do we bring the team
Speaker:together?
Speaker:When is it okay to let off steam?
Speaker:It's not just tactically,
Speaker:how do I make a phone call and
Speaker:do a pitch or this deal is stuck?
Speaker:How do I unstick it?
Speaker:I think those things we should
Speaker:outsource, but we shouldn't
Speaker:outsource the emotional care
Speaker:of our employees.
Speaker:So the coaching that we're referring
Speaker:to is really technical skills,
Speaker:competence coaching, as opposed to
Speaker:more of the employee broadly
Speaker:speaking around their
Speaker:life and their concerns and their
Speaker:kind of challenges that they're
Speaker:having, is your point.
Speaker:And I think that there will be
Speaker:people who want to do that because
Speaker:you have people that are using
Speaker:ChatGPT as a therapist.
Speaker:You have people who are bonding and
Speaker:having it be their best
Speaker:friend.
Speaker:But in a work environment,
Speaker:I think it's dangerous to lose
Speaker:the human connection.
Speaker:And even if difficult conversations
Speaker:and forming bonds with
Speaker:humans is something that not all
Speaker:managers want to, I think is
Speaker:important for a business
Speaker:to work well for those human bonds
Speaker:to still be there.
Speaker:And then the second point is
Speaker:around managing
Speaker:AI employees.
Speaker:And it's one of the things I've been
Speaker:thinking about, and I'm going to do
Speaker:some training and some process
Speaker:at work because basically
Speaker:everybody's going to have to become
Speaker:a manager. And not everybody has
Speaker:managerial experience.
Speaker:I was listening to a podcast
Speaker:where, I think,
Speaker:I can't remember, Deloitte
Speaker:Accenture, somebody had
Speaker:done a thing about AI
Speaker:adoption within the enterprise.
Speaker:It was like execs
Speaker:had the highest adoption, then
Speaker:middle management the next and then
Speaker:junior people the least.
Speaker:And like part of the methodology is
Speaker:you don't know if junior people
Speaker:are genuinely not using it or
Speaker:lying about not using because they
Speaker:don't what they should use and
Speaker:shouldn't use and so it's just
Speaker:safer to say they're not.
Speaker:So there's probably an element of
Speaker:that But then there's also an
Speaker:element of senior
Speaker:people and managers.
Speaker:Understand the art of delegation,
Speaker:understand what can be delegated,
Speaker:are very good at giving precise
Speaker:communication.
Speaker:So part of it is people not
Speaker:owning up, but the other part of it
Speaker:is I think that it's actually easier
Speaker:for senior execs and
Speaker:managers to adopt AI
Speaker:because it's a lot of the same
Speaker:managerial skills.
Speaker:You understand what you
Speaker:can delegate, what you cannot
Speaker:delegate, very good at giving
Speaker:precise.
Speaker:Instruction, very good at giving
Speaker:feedback when the instructions
Speaker:aren't precise, good at asking
Speaker:clarifying questions, and
Speaker:those are all the things that you do
Speaker:in order to get good results out of
Speaker:ChatGPT.
Speaker:And so one of the next
Speaker:things that we're going to do at
Speaker:Matomic is a
Speaker:training session for everyone on
Speaker:how to think about delegation,
Speaker:how to give precise questions.
Speaker:One of the great things about AI
Speaker:is it doesn't have feelings, so you
Speaker:don't have to do like.
Speaker:Was really good, but why
Speaker:don't you think about this instead
Speaker:or in addition to what you're doing?
Speaker:You can just be like, no, I said
Speaker:this.
Speaker:Why do you keep doing this?
Speaker:Walk me through your reasoning and
Speaker:it'll explain why it's doing the
Speaker:wrong thing. You're like, ah, no.
Speaker:This is what I asked and this is
Speaker:what i mean and do this.
Speaker:So I think this is a fabulous point
Speaker:and a point I haven't really heard
Speaker:before, which is, you're right,
Speaker:every single employee in the
Speaker:company, regardless of their level,
Speaker:has to be a manager to be able to
Speaker:manage AI employees.
Speaker:And not everybody's had that
Speaker:experience yet, so that's part of
Speaker:training that I'll be doing rather
Speaker:than bringing somebody in.
Speaker:And then the next stage after that
Speaker:is to think through the way that
Speaker:Matt has performance
Speaker:reviews, OKRs,
Speaker:job descriptions for
Speaker:your AI employees, so that you
Speaker:actually have a high-performance
Speaker:team rather than a
Speaker:team that goes a bit wonky.
Speaker:I'm thinking back to our previous
Speaker:conversations before you entered
Speaker:Atomic and kind of this AI
Speaker:chats that we had had.
Speaker:And I feel like between Matt,
Speaker:the inspiration of Matt and now what
Speaker:you've actually picked up here, if
Speaker:you're thinking about an entry
Speaker:point in your company of
Speaker:where to start with this stuff,
Speaker:where you're moving past just using
Speaker:individually, just using chat GBT
Speaker:as a bit of a support mechanism for
Speaker:one-off prompting, which we're all
Speaker:doing right now, but really the
Speaker:question as an operator of how you
Speaker:take this and do something that is
Speaker:much more.
Speaker:Durable and evolved within
Speaker:an organization to really transform
Speaker:it as opposed to this one-off stuff
Speaker:that we're currently doing.
Speaker:This entry point of the mindset
Speaker:and the idea that these are AI
Speaker:employees that need to be onboarded,
Speaker:contextualized, measured,
Speaker:and so on is maybe the key
Speaker:mindset distinction of the actual
Speaker:way in which you can take this AI
Speaker:stuff in a company right now and
Speaker:set yourself on a journey.
Speaker:Yeah, 100%. And I
Speaker:feel like it's all over LinkedIn,
Speaker:and we might be a bit late.
Speaker:But then in reality, when I speak to
Speaker:people, nobody's even like
Speaker:my husband has to keep reminding me
Speaker:that I'm an early adopter.
Speaker:Yeah, I think there's so I mean I
Speaker:might be deluding myself here, but I
Speaker:feel like you are ahead of the game
Speaker:I think, there's very few mats that
Speaker:are out there that are doing this
Speaker:stuff in a very earnest way
Speaker:I hope so because this is going to
Speaker:be our competitive advantage and now
Speaker:sharing it with everybody else, but
Speaker:you know, it's around how you
Speaker:actually do it rather than just the
Speaker:idea.
Speaker:So let's move on to our conversation
Speaker:with Matt Lummo,
Speaker:AI employees.
Speaker:Have you seen an increase
Speaker:in productivity per AE?
Speaker:And by productivity, I mean how much
Speaker:they are selling, like how much ARR
Speaker:they're selling per head.
Speaker:Yes, just to give you some numbers,
Speaker:but basically I gave the goal to
Speaker:all our departments in the company
Speaker:to be able to do at least 3x
Speaker:productivity over Q1 and
Speaker:Q2 of 2025.
Speaker:We reached the 3x productivity
Speaker:by March basically.
Speaker:So right now, whether it's on the
Speaker:sales side or on the
Speaker:developers, for instance, we're more
Speaker:in the 5 to 6x productivity.
Speaker:And currently what we're doing is
Speaker:we're actually starting to hire a
Speaker:bunch of AI employees.
Speaker:And we're really calling them this
Speaker:way, we actually have a job
Speaker:description and we even give them a
Speaker:name. And so this is important
Speaker:because we're trying right now to
Speaker:really let our company not
Speaker:start using AI as a tool,
Speaker:but more to partner with AI.
Speaker:And I think this is an important
Speaker:distinction that people need to
Speaker:start making because AI is not
Speaker:a tool. If you're just using a tool
Speaker:and you expect to have the result
Speaker:right away, it's not going
Speaker:to work.
Speaker:If you want to use AI well,
Speaker:you have to consider AI as
Speaker:an employee, meaning that you
Speaker:can hire the best person on the
Speaker:market if you need to onboard that
Speaker:person.
Speaker:You need to make sure that person
Speaker:has all the context of your company,
Speaker:of your deal, of your offering.
Speaker:And then you need do performance
Speaker:reviews with that person, making
Speaker:sure that they're actually
Speaker:delivering on the job.
Speaker:And if not, making changes in
Speaker:the instructions, et cetera.
Speaker:And so it's interesting because
Speaker:since we started this initiative,
Speaker:we've seen that the use of AI has
Speaker:changed, the
Speaker:understanding of it has changed, and
Speaker:we're starting to get better and
Speaker:better and the results of this.
Speaker:And we're just at the beginning of
Speaker:it, right? So my personal goal
Speaker:right now is by the end of this
Speaker:year, I was studying you were on the
Speaker:5x to 6x right now, depending on the
Speaker:teams. I think we'll be at 10x
Speaker:compared to the beginning of the
Speaker:year as we start wrapping up more
Speaker:and more with new AI colleagues for
Speaker:everyone.
Speaker:What tools are you using?
Speaker:Basically just cloud and chat GPT.
Speaker:And then any automation like
Speaker:clay or N8N
Speaker:or Zapier.
Speaker:Yeah, we have a lot of things behind
Speaker:this, but what's interesting is we
Speaker:actually, the NA8 and the others,
Speaker:we actually use them less and less
Speaker:because obviously the capacities of
Speaker:Shared GPT and Cloud have improved
Speaker:more and more.
Speaker:You are able now to have subagents
Speaker:that can run.
Speaker:So there is a lot things you can
Speaker:start doing natively.
Speaker:And I would expect, honestly, I
Speaker:think in August SharedGPT
Speaker:5 will be released.
Speaker:And I think we're probably going to
Speaker:see more and more ability for
Speaker:everyone to just do everything with
Speaker:directly in their app.
Speaker:How do you create an employee?
Speaker:Who creates an employee and then
Speaker:manages them?
Speaker:Well, it's very easy.
Speaker:So it basically it's whoever think
Speaker:about it, just a normal employee.
Speaker:So if you have someone is in charge
Speaker:of marketing, the marketing leader
Speaker:is going to say, well, who would
Speaker:hire? Let's say I have unlimited
Speaker:budget and I can hire as many people
Speaker:as I want.
Speaker:What do I need?
Speaker:And it's very important to think
Speaker:this way and not thinking that with
Speaker:one AI tool, you can solve all the
Speaker:problems. No.
Speaker:Well, for the case of marketing, for
Speaker:instance, well, you probably want
Speaker:someone to write the content of your
Speaker:website. You also want someone to
Speaker:write your content of emails,
Speaker:because it's not the same type of
Speaker:audience. It's not the same format.
Speaker:And when it comes to emails,
Speaker:probably outbound versus
Speaker:your typical marketing emails, it's
Speaker:probably not the skill set, right?
Speaker:And so you just want basically to
Speaker:divide these roles, this way,
Speaker:build a job description for them.
Speaker:Then you need to basically feed the
Speaker:AI with what we call internally
Speaker:blueprints, which is all the context
Speaker:of your company.
Speaker:So you're offering your customers,
Speaker:your segmentation, etc.
Speaker:And then you basically have to
Speaker:be a bit smart about the prompts, a
Speaker:lot of iterations.
Speaker:But eventually you get to a point
Speaker:where we can build content, for
Speaker:instance, for a new webpage with
Speaker:just asking our teammates.
Speaker:So, you know, we have Emily
Speaker:and we tell Emily like, well, I need
Speaker:a new page for this feature we're
Speaker:launching. Let's about it.
Speaker:And Emily just dropped the page.
Speaker:And if she has questions, she will
Speaker:ask us.
Speaker:It's like a conversation with a
Speaker:colleague.
Speaker:And you'll be surprised by the
Speaker:quality of the results when you
Speaker:really spend time onboarding
Speaker:well your AI tool.
Speaker:Bethany was just talking about the
Speaker:fact that product marketing is going
Speaker:to be replaced by this, essentially,
Speaker:and my argument back to her was that
Speaker:I think there's always going to be
Speaker:that human top-up of it's
Speaker:going to It always gets to a point
Speaker:where it's still not quite fit for
Speaker:purpose basically and somebody has
Speaker:to look at it with human eyes to
Speaker:really right size it for whatever
Speaker:purpose that you have and Bethany's
Speaker:response to that was well, that's
Speaker:probably not going to be the case
Speaker:going forward, Brandon.
Speaker:I need to hire Emily, the product
Speaker:marketer here in ChatGBT.
Speaker:Here's a tremendous amount of
Speaker:context, blueprints, whatever you
Speaker:want to call it to get them in a
Speaker:space where they're highly targeted.
Speaker:And then feed them your request
Speaker:that you have for product marketing
Speaker:purposes, give me copy for the
Speaker:website or position or whatever I
Speaker:guess in that case, that's
Speaker:essentially what you're doing.
Speaker:Exactly.
Speaker:And I think, you know, I think
Speaker:product marketers, many rules like
Speaker:that are in danger.
Speaker:But I think what does that mean for
Speaker:them? For if someone tomorrow is a
Speaker:product marketer, what should you be
Speaker:doing? What you should be doing is
Speaker:really understanding how to build
Speaker:these.
Speaker:AI colleagues, because this way
Speaker:you own the system.
Speaker:And you will always need someone,
Speaker:for now, I hope for the next five or
Speaker:10 years, to actually build that
Speaker:system because you do need to
Speaker:onboard that employee.
Speaker:You do need just to give them the
Speaker:philosophy of your company when it
Speaker:comes to the voice and things like
Speaker:this. It's not something that the AI
Speaker:should choose.
Speaker:We still have human beings to do
Speaker:this. And I kind of tell my
Speaker:teams this is basically today now,
Speaker:as they start working with these AI
Speaker:colleagues they're basically their
Speaker:own little VP.
Speaker:It's really what it is. Have a team
Speaker:of sometimes five, six, ten
Speaker:colleagues that will be able to run
Speaker:a different part of their
Speaker:work, but they still have to
Speaker:coordinate all of this to make sure
Speaker:that the work is done according to
Speaker:the standards of what we want.
Speaker:And so they become managers, it's
Speaker:just not real people, it is AI
Speaker:colleagues.
Speaker:And so are you using projects for
Speaker:each, like each colleague is a
Speaker:project in effect?
Speaker:Exactly. Yeah, that's the way we
Speaker:use it in Cloud in particular, I
Speaker:guess.
Speaker:And are you using Gemini?
Speaker:Because Gemini was so bad, I
Speaker:didn't use it. And now I tried it
Speaker:the other day and it was actually
Speaker:quite good.
Speaker:We use Gemini Mirror on the product
Speaker:side, not on the go-to-market
Speaker:teams.
Speaker:It's changing, and probably Gemini
Speaker:is going to be out of that very,
Speaker:very soon. So I think we'll see some
Speaker:progress. What's really exciting is,
Speaker:honestly, we've launched that
Speaker:recently, right?
Speaker:But we're also doing this because we
Speaker:know that within a month or two, new
Speaker:models are coming out.
Speaker:There is new options all the time
Speaker:and more things, right.
Speaker:GPD just released a few weeks ago
Speaker:the new improvements for
Speaker:Operator.
Speaker:And so if things are working
Speaker:today relatively well for us,
Speaker:I am absolutely certain that within
Speaker:a few months it would be probably
Speaker:even better.
Speaker:And so I think it's a question of
Speaker:jumping now on this because
Speaker:it does require a lot of work to
Speaker:organize your context to really
Speaker:understand how to interact with the
Speaker:AI. But if you do it now,
Speaker:then when models are getting better,
Speaker:you're going to be ahead of
Speaker:everyone. And that's kind of what
Speaker:we're trying to do.
Speaker:But salespeople still need to close
Speaker:deals, right?
Speaker:They do, and they should keep doing
Speaker:this, but they should not spend time
Speaker:anymore sending emails and
Speaker:follow-ups.
Speaker:They should not taking any more time
Speaker:to review an account or do discovery
Speaker:before a call,
Speaker:and this should not waste time in a
Speaker:meeting with a manager coaching them
Speaker:on their calls.
Speaker:That's the type of thing that they
Speaker:should avoid.
Speaker:I think the way we look at this is
Speaker:whether it's our customer success
Speaker:team or sales team, they should
Speaker:spend most of their time.
Speaker:And when I say most, it's 80% of
Speaker:their Thank you for your time.
Speaker:In touch with customers, whether
Speaker:it's on a call, for a chat,
Speaker:there's different ways to interact
Speaker:with them, right?
Speaker:But that's the value of a human
Speaker:being.
Speaker:It's the relationship that you
Speaker:create.
Speaker:And that's one thing that I think AI
Speaker:isn't going to replace yet.
Speaker:We'll see in a few years, right, but
Speaker:I think for now we're safe.
Speaker:Yeah, sometimes we're safe.
Speaker:And then I was listening to a
Speaker:podcast today where they were
Speaker:talking about chat GPT
Speaker:therapists and how they're actually
Speaker:fairly good.
Speaker:So who knows?
Speaker:Well, the thing is your therapists
Speaker:are like AI, right?
Speaker:If you interact the wrong way with
Speaker:them, you're not gonna get what you
Speaker:want. If at the end of the day, you
Speaker:go see a therapist just because you
Speaker:want to hear someone telling you
Speaker:like, that must be hard for you.
Speaker:Well, sure, but you can also just
Speaker:talk to your grandmother or to a GPT
Speaker:and you'll have a lot of the same results.
Speaker:If you want someone to challenge
Speaker:you, then you have to find the right
Speaker:therapist. And that's the thing with
Speaker:AI. You need to tell the AI that you
Speaker:wanna be challenged, not, you know,
Speaker:comforted.
Speaker:If our listeners can only
Speaker:take one thing away from
Speaker:today's conversation,
Speaker:what would that one thing be?
Speaker:What we just talked about AI
Speaker:employees.
Speaker:I think the misunderstanding still
Speaker:today in companies and
Speaker:sometimes tech companies about
Speaker:what AI can really do.
Speaker:I think there is a significant gap
Speaker:right now. And I think that gap is
Speaker:growing by the day because you're
Speaker:going to have some companies that
Speaker:understand this, some companies that
Speaker:do not.
Speaker:And I, think we are going to
Speaker:see a lot of replacements
Speaker:because once you miss the
Speaker:train, you're not able to catch up.
Speaker:And so that's the one thing I would
Speaker:really encourage everyone to look
Speaker:into is AI is
Speaker:10 times more powerful than what
Speaker:people think it is today.
Speaker:And there is a lot of things
Speaker:coming and it's
Speaker:time to really change and
Speaker:adapt all the processes and the way
Speaker:we work.
Speaker:So on that note, thank you, Matt,
Speaker:for joining us in the operations
Speaker:room. If you like what you hear,
Speaker:please comment or subscribe and we
Speaker:will see you next week.
