Fireside Chat with Mark Roberge — The Science of Agentic Scaling

385
01:00:55.080 –> 01:01:01.790
Julia Nimchinski: This is amazing. We are transitioning to our session on the science of agentic scaling with Mark Roberish and many.

386
01:01:01.790 –> 01:01:03.500
Manny Medina: Oh, Marco Bears.

387
01:01:03.500 –> 01:01:05.219
Julia Nimchinski: Two founders, two CEOs.

388
01:01:05.220 –> 01:01:08.009
Mark Roberge: Come on, Mandy. What’s up, man? How are ya?

389
01:01:08.910 –> 01:01:11.710
Mark Roberge: Dude, I’m bringing the HBS students out in January.

390
01:01:11.710 –> 01:01:12.429
Manny Medina: I know that.

391
01:01:12.430 –> 01:01:14.759
Mark Roberge: We can go out your way.

392
01:01:14.760 –> 01:01:16.060
Manny Medina: This professorship is…

393
01:01:16.060 –> 01:01:21.519
Mark Roberge: Dude, these kids actually want to learn how to sell now. Hey, Mark, how are you?

394
01:01:21.520 –> 01:01:22.579
Mark Organ: Hey, good to see.

395
01:01:22.580 –> 01:01:27.260
Mark Roberge: I think the last time I was with you was we were in, like, Paris or something. I don’t remember where we did that.

396
01:01:27.260 –> 01:01:27.939
Mark Organ: We’re at it.

397
01:01:27.940 –> 01:01:28.490
Manny Medina: Everywhere.

398
01:01:28.490 –> 01:01:32.680
Mark Roberge: It was a speech, man, it wasn’t fun, it was in and out.

399
01:01:32.680 –> 01:01:36.659
Mark Organ: I look forward to seeing you again, maybe at SAS Talk in Ireland in a couple weeks.

400
01:01:36.820 –> 01:01:37.980
Mark Roberge: Yeah, that’d be awesome, man.

401
01:01:38.110 –> 01:01:38.830
Mark Organ: Cheers!

402
01:01:39.670 –> 01:01:40.380
Manny Medina: Yeah, guys.

403
01:01:41.500 –> 01:01:44.130
Julia Nimchinski: Awesome, welcome to the show! Mark.

404
01:01:44.130 –> 01:01:52.790
Mark Roberge: Julia, you always put a great one on, and loved the questions to prep. It’s you’re always on the edge, so I’m excited to chat.

405
01:01:53.480 –> 01:02:01.029
Julia Nimchinski: Always my pleasure, and I don’t know, to everyone who doesn’t know you, if there’s anyone watching.

406
01:02:01.290 –> 01:02:05.959
Julia Nimchinski: But yeah, Mark is the founding CRO, HubSpot.

407
01:02:06.180 –> 01:02:16.469
Julia Nimchinski: co-founder at Stage 2 Capital, and a professor at the Harvard Business School. Super excited to dive in into our fireside chat, The Science of Agentech Scaling.

408
01:02:16.660 –> 01:02:23.429
Julia Nimchinski: And, yeah, I’ve prepared, I think, questions… I guess it’s gonna be enough for 5 fireside chats, or all.

409
01:02:23.430 –> 01:02:29.849
Mark Roberge: I don’t know, I don’t think we’re gonna get through the first one. There’s a lot in there. We’re definitely gonna get… but there were some good ones.

410
01:02:29.850 –> 01:02:46.959
Julia Nimchinski: Let’s do it. So the first one, first of all, I can’t wait when you will release the book, The Signs of Scaling, but if you were to write it all over again, what chapter would you add for founders or builders in the Agent Tech era?

411
01:02:47.470 –> 01:02:58.009
Mark Roberge: Yeah, it’s, so I think that was related to the first book 10 years ago, the Sales Acceleration Formula. The… I did talk… think a lot about the science of scaling principles.

412
01:02:58.550 –> 01:03:10.550
Mark Roberge: which have been the born of the last 10 years of research and working with companies on scale, and many of the under, I think almost all of the attributes, and we’ll dig into it, apply as we go into a very aggressive agentic era.

413
01:03:10.630 –> 01:03:22.640
Mark Roberge: But I did have to create an appendix, to, like, explain how it might work. So, the first framing, Julia, is… I think it’s really difficult to…

414
01:03:22.830 –> 01:03:35.219
Mark Roberge: predict the timing and extent of abstraction from today’s point of work. So let me just explain that, like, I… I use, like, a building a toy example. You’re a company, you build toys.

415
01:03:35.610 –> 01:03:46.499
Mark Roberge: I think the first abstraction would be, like, you use AI to support you through the stages of building a toy. Product research, customer research, manufacturing planning, etc, etc.

416
01:03:46.830 –> 01:03:49.209
Mark Roberge: The second abstraction would be, like.

417
01:03:49.430 –> 01:03:56.450
Mark Roberge: just build the toy. You know, like, just go and just figure out how to build the toy. And the third obstruction might be, like.

418
01:03:56.510 –> 01:04:14.260
Mark Roberge: well, what’s the point of a toy? The toy is to, like, create enjoyment for children and create a profit as a company, so, like, create a way to create enjoyment for children. You know, so you get the point, like, it’s just… it’s really difficult to figure out the timing of all those things and how far we actually go. And so, when I think about that on the go-to-market side.

419
01:04:14.610 –> 01:04:29.479
Mark Roberge: I’ll propose, like, 4 levels of abstraction. I think the first one we’re actually going through right now, which is to, basically, the seller still… a human seller still sells to a human buyer.

420
01:04:29.530 –> 01:04:31.650
Mark Roberge: But all the other stuff goes away.

421
01:04:31.680 –> 01:04:44.660
Mark Roberge: Y’all, so we go, I think the average selling time right now, the average amount of time that a salesperson spends in front of a customer is around 25% across all companies, huge volatility. That can go to, like, 90%.

422
01:04:44.660 –> 01:04:51.800
Mark Roberge: And so that would be the first phase, is they do all, like, the support mechanism. Phase two is the… the seller is actually an agent.

423
01:04:51.970 –> 01:05:05.919
Mark Roberge: Phase 3 is the buyer’s an agent, and I know you had Carol on, and she’s kind of studying that very closely. And then Phase 3 is, like, the concept of selling gets integrated in all the other functions of, like, product development, and now, like.

424
01:05:06.030 –> 01:05:11.339
Mark Roberge: functional specialization goes away, and it’s all, like… and so…

425
01:05:11.530 –> 01:05:15.219
Mark Roberge: Within that abstract… does that make sense, Julia?

426
01:05:15.220 –> 01:05:15.950
Julia Nimchinski: Absolutely.

427
01:05:15.950 –> 01:05:29.679
Mark Roberge: what I’m talking about? Okay, cool. So I’ll just talk about, like, the first one that, I think we’re… we haven’t… I think 2024 was sort of the year of, like, breakouts in product development with AI, so it’s huge.

428
01:05:29.870 –> 01:05:35.470
Mark Roberge: up-leveling in terms of efficiency. I don’t think we’ve seen it go to market yet. I can’t wait.

429
01:05:35.990 –> 01:05:41.339
Mark Roberge: till we have, like, the first rep that, like, says, in 2024, I…

430
01:05:41.530 –> 01:05:45.340
Mark Roberge: did a million in revenue, and this year I’m gonna do 10 million.

431
01:05:45.580 –> 01:05:51.239
Mark Roberge: Have you seen that yet? Because of AI, not because they closed some huge deal. Have you seen that? I haven’t seen that yet.

432
01:05:51.930 –> 01:06:02.149
Julia Nimchinski: there are a lot of claims of three type of team, I don’t know, founders hitting 30 million ARR. Oh yeah, that’s out there. Yeah, I mean, it’s like…

433
01:06:02.150 –> 01:06:08.880
Mark Roberge: I don’t… I… I haven’t seen anything that gets me excited yet on that. They’re doing it, absolutely.

434
01:06:09.710 –> 01:06:17.090
Mark Roberge: But I wish when I… double-clicked into those stories, I saw that, like, they…

435
01:06:17.200 –> 01:06:32.649
Mark Roberge: were just, like, the humans were 10x more productive, that the customers had a lot of, like, production, mission-critical implementations of the AI, that churn was good, you know, strong. It’s just like…

436
01:06:33.370 –> 01:06:39.660
Mark Roberge: For the most part, these are… they’re not sales-driven, they’re PLG-driven, which is great.

437
01:06:39.800 –> 01:06:49.279
Mark Roberge: But I feel like a lot of the adoption is in experiments, innovation labs. They’re not, like, mission critical yet.

438
01:06:49.280 –> 01:07:02.729
Mark Roberge: churn’s kind of high on some of them, the usage isn’t exceptional. And I think the other big concern I have in there is the flywheel that’s occurring within the hype cycle. Like, I think some of these have

439
01:07:03.120 –> 01:07:22.150
Mark Roberge: a lot of their install base and revenue comes from other AI-native companies, and that’s really concerning to me, because that’s what happened in the dot-com era, is, like, you were… you were doing well, but a lot of your revenue came from other internet startups, and when it burst, like, everything fell apart, and then everybody can remember 2022,

440
01:07:22.150 –> 01:07:25.449
Mark Roberge: When so much of the tech bubble happened.

441
01:07:25.450 –> 01:07:42.789
Mark Roberge: and a lot of the customer base with other tech companies that were growing like crazy, and when that fell apart, everything fell apart. So, I just… the jury’s out on whether we’re in a hype cycle or not in AI. I have a lot of conviction we are, and those are some of the things I, like, tend to look at, and…

442
01:07:42.790 –> 01:07:45.839
Mark Roberge: get concerned about. I think we will have, like, the…

443
01:07:46.010 –> 01:08:00.560
Mark Roberge: $100 million business with 3 companies, 3 employees, but I don’t think the… today’s implementation is so much that. But just to finish your question, Julie, on, like, predictions on go-to-market, because I do think, like.

444
01:08:00.560 –> 01:08:11.139
Mark Roberge: In the next year, we will see a massive shift in… I’ll summarize it as our selling time will go from 25% to, like, 80%, our customer-facing time.

445
01:08:11.260 –> 01:08:21.669
Mark Roberge: And I think, like, you can break down each element of the customer journey and the support process needed in sales. So let’s just talk with ICP identification, right? Like.

446
01:08:21.899 –> 01:08:24.980
Mark Roberge: Right now, we, like, look at various segments.

447
01:08:25.430 –> 01:08:36.440
Mark Roberge: We, if you do it right, you calculate your ICP based on customer success creation and retention. You make sure that you can have, like, strong go-to-market CAC around it as well.

448
01:08:36.560 –> 01:08:40.629
Mark Roberge: And… there’s just a lot of work. We probably refresh it every quarter.

449
01:08:40.689 –> 01:08:59.270
Mark Roberge: you know, our… and, like, that’s something that can be completely automated and done in real time. Like, every day, you could be like, oh, guess what? There’s, like, 20 new companies that are in our ICP now, because our ICP’s expanded. And then that bleeds into, like, account identification. Like, most companies to this day.

450
01:09:00.069 –> 01:09:02.870
Mark Roberge: They’re… that’s a huge decision, like…

451
01:09:03.069 –> 01:09:07.540
Mark Roberge: You can… you have sales capacity to call 70 new accounts this month.

452
01:09:08.609 –> 01:09:13.880
Mark Roberge: And that decision is being put in a 22-year-old SDR. That’s like…

453
01:09:14.250 –> 01:09:30.190
Mark Roberge: I mean, a lot of guidance is great, but that’s an easy thing for AI to, like, now that we know our ICP, we can identify, like, the… here’s 70, here are the… if you have time to go after 70 new accounts, this is, like, the perfectly statistically optimized 70 accounts you should go after.

454
01:09:30.290 –> 01:09:46.880
Mark Roberge: And then the third one is, like, going after them. Who are the different members of the decision-making unit? What is the optimal outreach to them? Like, we’re starting to see that, come to play. And, like, we’ll see, like, how much of that’s delivered through SDRs or not, whatever, we’ll get there.

455
01:09:46.920 –> 01:09:56.450
Mark Roberge: And then I think that we’ll have a huge enablement era, too. Like, we’re already seeing… there’s an awesome company called Letter AI that’s, like, working on this right now, where…

456
01:09:56.590 –> 01:10:04.970
Mark Roberge: you’re a rep, and you have a meeting tomorrow. You just booked your first meeting with, like, the head of… director of marketing at Genzyme.

457
01:10:05.600 –> 01:10:08.590
Mark Roberge: And… It basically…

458
01:10:08.610 –> 01:10:27.070
Mark Roberge: has already been pre-trained on your whole product, your battle cards, your sales methodology. It looks at, you know, Mary from Genzyme, and evaluates… reads everything about Genzyme, and reads everything about Mary’s background, and basically says, this is the perfect first meeting in terms of, like, discovery call questions, etc.

459
01:10:27.070 –> 01:10:39.580
Mark Roberge: It spins up a buyer agent of Mary that you can practice with the night before, so you don’t have to wait for your trainer or your manager to do that, you can just do it, it coaches you on it, and then it comes to the meeting with you, whether it’s, like, a Slack

460
01:10:39.630 –> 01:10:45.140
Mark Roberge: bought on the side, or it’s actually there with you. And as information’s disclosed, it…

461
01:10:45.210 –> 01:10:53.709
Mark Roberge: adopts our knowledge and proposes, like, stories or questions to ask. And then after the meeting, obviously, it automatically

462
01:10:53.830 –> 01:11:05.360
Mark Roberge: fills out the CRM, updates the forecast, like, decides, like, creates a deal room, and, like, whatever the next step should be, and also updates my skill scorecard of, like.

463
01:11:05.360 –> 01:11:16.700
Mark Roberge: Hey, Mark, by the way, next time you’re in, like, a 7-minute Uber ride, like, this is the video you should take, because you’re still only scoring a 3.7 on 10 on sense of urgency Development. Like, that’s all, like, happening now, right? So…

464
01:11:16.720 –> 01:11:20.430
Mark Roberge: And then that bleeds into the forecast, model.

465
01:11:20.530 –> 01:11:23.590
Mark Roberge: Where it’s like, I mean, that’s just…

466
01:11:24.400 –> 01:11:30.920
Mark Roberge: You shouldn’t have… it’s just the transcripts of discussions with customers, and that generates the forecast.

467
01:11:31.010 –> 01:11:48.909
Mark Roberge: And, like, if you look at, like, Momentum AI as an example, they’re doing great stuff with this. So those would be the elements that I think are gonna happen in the next 12 months, and they are happening. People are exploiting them today, is agent… whether this is a single agent or different agents, like ICP identification.

468
01:11:48.910 –> 01:12:03.870
Mark Roberge: New target account creation, individual, you know, decision-making unit, identification, and demand gen, preparing and supporting through the meeting, updating the CRM and the forecast, and skills update. And I forgot the last one, which was hiring.

469
01:12:04.000 –> 01:12:20.110
Mark Roberge: Because hiring is just such another… if you’re good every quarter, you go back to the hires you made 6 or 9 months ago. Are they doing well, and are they not, and did we catch it, or did we not, and did we update the scorecard? That’s a perfect learning model that can be done with AI. So those are the…

470
01:12:20.110 –> 01:12:35.890
Mark Roberge: the things that are happening, I think, in the next year. And if Cara Lou’s right, and next summer we have buyer agents, that would be cool. I love Carolu. I think she’s one of the most, brilliant thought leaders in marketing. I’ll bet against that. I’ll do 100 bucks.

471
01:12:37.000 –> 01:12:52.360
Mark Roberge: Guys, I just think, like, it always takes longer. Like, I just… especially San Francisco, like, they always think it’s coming faster. It’s great, I mean, the best thought leadership in the world. But I just think, like, it’s gonna come, but I think it’ll come a little slower.

472
01:12:53.920 –> 01:12:57.170
Julia Nimchinski: Mark, building on this, so you mentioned coaching.

473
01:12:57.780 –> 01:13:05.689
Julia Nimchinski: I don’t know, coaching agents, be it. There are a lot of founders that are innovating in this space, but what are your thoughts on the adoption?

474
01:13:06.060 –> 01:13:09.530
Julia Nimchinski: And how do you actually, I don’t know, enforce it? Is it…

475
01:13:09.890 –> 01:13:18.349
Julia Nimchinski: Do you see any changing… changes in leading indicators, lagging indicators, when it comes to retention and the adoption of those agents?

476
01:13:18.990 –> 01:13:24.649
Mark Roberge: So coaching reps? Yeah. Coaching sales reps? Yeah, it’s coming. You know, I think we had a couple, like.

477
01:13:24.850 –> 01:13:30.730
Mark Roberge: early movers that the product didn’t work great, and that’s another, like, concern I have in…

478
01:13:30.900 –> 01:13:42.420
Mark Roberge: Like, right now, hype cycle is, it’s just, like, absolutely rigorously researched, statistically proven that we, as an investor and entrepreneur community.

479
01:13:42.420 –> 01:14:05.259
Mark Roberge: over-emphasize first mover advantage. It’s just, like, there’s millions of studies, and, like, pick your study. Roughly speaking, the first mover wins 12% of the time, and the fast follower wins 88% of the time. And so, like… and it makes sense. Like, the first mover has to, like, deal with the primitive tech, and waste all this money on product development, and evangelize the category, and…

480
01:14:05.260 –> 01:14:25.729
Mark Roberge: get the pricing wrong, and, like, all that shit. And all the fast follower needs to do is, like, once they’ve named the category and built the product, they just have to re-engineer it, underprice it, and benefit from all that with the best… you know, it’s just like, there’s a long reason why that’s the case, and I think we’re just working our way through, like, the first movers on the coaching products that just didn’t… they flopped.

481
01:14:25.730 –> 01:14:27.770
Mark Roberge: But we’re, we’re getting there.

482
01:14:27.770 –> 01:14:40.049
Mark Roberge: And I think with any, if I were driving that internally as a CRO, just don’t try to do it with everyone. You always have to start with the early adopters, you know, who are probably gonna be…

483
01:14:40.050 –> 01:14:58.629
Mark Roberge: your newer hires, they’re probably gonna be the people that are early in their career, they’re, like, their appetite for learning is massive, and just focus on them. Just, like, pick two, three, four of them, focus on them. And then, as they start to dramatically improve, you can get more aggressive around, forcing it, through the laggards.

484
01:14:59.620 –> 01:15:09.190
Julia Nimchinski: So in both of your books, you’re presented almost like a qualification matrix, and there is a lot of, on, on hiring,

485
01:15:09.760 –> 01:15:15.360
Julia Nimchinski: Generally, salespeople and customers. And there was a lot of emphasis on people.

486
01:15:15.490 –> 01:15:24.719
Julia Nimchinski: And I get it, you’re saying that we are not into, like, I don’t know, we’re not gonna transition into buying bots by 2026, midsummer, whatever.

487
01:15:24.990 –> 01:15:38.119
Julia Nimchinski: But since this transition is going to happen, how do you see agents in this equation? Are we moving towards agent market fit? I don’t know, product agent fit?

488
01:15:38.840 –> 01:15:47.450
Mark Roberge: Yeah. Yeah, it’s a great question. So, just to, like, back up and form the foundation for the audience, like, the premise of my most recent work

489
01:15:47.910 –> 01:15:56.879
Mark Roberge: is called The Science of Scaling, which, over the last 10, 15 years, I’ve been… I run a VC firm, we’ve made probably 100 investments. Before that, I…

490
01:15:57.040 –> 01:16:04.040
Mark Roberge: Spent a day a week with, like, 20 different companies to, like, really dig into their go-to-market motion.

491
01:16:04.040 –> 01:16:13.520
Mark Roberge: And, had an opportunity to pattern recognize on failure versus success. And a lot of the… I just found a lot of unnecessary failure

492
01:16:13.520 –> 01:16:26.729
Mark Roberge: by a lack of rigor around deciding when to scale revenue and how fast. I just think it’s a lot of, like, like this. You know, it’s like, we just raised a round, scale fast. Hire 20 reps next month and see what happens. Like, it’s…

493
01:16:26.970 –> 01:16:46.349
Mark Roberge: Triple, triple, double, double, because that’s what Snowflake did. I don’t know, it’s just like a lot of… it’s a pretty important decision, and there’s no rigor. So, that’s really the concept of the science scaling, is it allows founders and board members to use the internal data of the company to decide and calculate when they should scale revenue and calculate how fast they’re capable.

494
01:16:46.350 –> 01:16:51.119
Mark Roberge: And it talks about 3 phases of you have to first find product-market fit.

495
01:16:51.120 –> 01:17:13.610
Mark Roberge: Which, you know, of course, dah, but like, hey, ask 100 entrepreneurs what product market fit is, you get 100 different answers, and half of them are around a revenue target, which I wholeheartedly disagree with. That just means you can sell. So I just think the premise of product market fit is, in our industry, Julie, is about retention, and the lead indicator of retention as indicated by creating customer success. So it talks a lot about that.

496
01:17:13.610 –> 01:17:19.979
Mark Roberge: Once you have that, you go to go-to-market fit, which means, okay, cool, you’re gonna sign up 10 customers this month.

497
01:17:20.030 –> 01:17:32.339
Mark Roberge: you’ve just proven that all of them are gonna be successful. You haven’t proven that you can build a business, i.e, acquire those customers at a low enough CAC to make the LTV formula work. So that’s go-to-market fit, and that’s when, like.

498
01:17:32.340 –> 01:17:50.170
Mark Roberge: Building playbooks, hiring reps, like, figure out your pricing model, figure out your comp plan, figure out your quota, that all comes into play. That shouldn’t be discussed during Product Market Fit. And then once you get that right, then you go into growth and mode, which you start adding, whatever, 2 reps a month, two reps a quarter, and there’s a whole other set of sequences that come out.

499
01:17:50.230 –> 01:18:03.660
Mark Roberge: So, what you’re talking about, Julia, is, you know, every element of the go-to-market system design changes as you navigate those stages. Your pricing model, your demand gen strategy, the people you hire.

500
01:18:04.190 –> 01:18:07.669
Mark Roberge: And you’re talking about the people, like, so, historically.

501
01:18:07.770 –> 01:18:14.120
Mark Roberge: The best salesperson at Salesforce.com right now would be a terrible first hire for a startup.

502
01:18:14.320 –> 01:18:15.349
Mark Roberge: It’s just like…

503
01:18:16.100 –> 01:18:21.739
Mark Roberge: I mean, the sales learning curve talks about this, like, I forget, the coin-operative rep. It’s like…

504
01:18:21.760 –> 01:18:38.559
Mark Roberge: that rep is awesome, because they came to Salesforce, went through a month of training, learned all the battle cards, learned the sales process, have a manager supporting them, had to… I mean, you don’t have any of that. You don’t even have product market fit. So, like, the first rep hire has to be, like, a PM plus an account executive, like…

505
01:18:38.570 –> 01:18:53.810
Mark Roberge: be able to go out and talk to 20 customers every week, whatever, and then sense-make and summarize the findings and talk to engineers about it so you can accelerate your journey. That’s just one example. Okay, so, to your point, Julia, how does this change in a gentic era? Well…

506
01:18:53.980 –> 01:18:57.950
Mark Roberge: You don’t need the product management sense-making aspect anymore.

507
01:18:58.170 –> 01:19:14.810
Mark Roberge: Because the rep should just go out and talk to as many customers as possible, and do what I’ll call the consultative selling, and the transcripts, the raw transcripts, can be used and analyzed to figure out where the product market fit is. So that skill goes away.

508
01:19:14.810 –> 01:19:18.179
Mark Roberge: Okay, so that’s one example of how this would change.

509
01:19:18.250 –> 01:19:20.970
Mark Roberge: The other thing I think that changes is…

510
01:19:21.660 –> 01:19:31.050
Mark Roberge: Scale can be more real-time with less upfront cost. So, today, it’s like, okay, cool, we find product-market fit.

511
01:19:31.210 –> 01:19:42.060
Mark Roberge: now let’s start scaling. So let’s start adding 2 reps a month. It’s gonna take us 60 days to find a rep. It’s gonna take us 4 months to ramp a rep.

512
01:19:43.010 –> 01:19:51.580
Mark Roberge: we’re gonna get it wrong 25% of the time. There’s just this, like, massive… Forward-looking latency of investment.

513
01:19:51.930 –> 01:19:57.609
Mark Roberge: And then, God forbid, we wake up with 5 reps and we don’t have a demand gen strategy to feed them, like…

514
01:19:57.780 –> 01:20:01.560
Mark Roberge: Versus, like, this becomes a lot more real-time in an agentic era.

515
01:20:01.670 –> 01:20:06.910
Mark Roberge: And… You can be going more, like, day-to-day as opposed to quarterly planning.

516
01:20:07.130 –> 01:20:10.190
Mark Roberge: And it also, because of that.

517
01:20:10.580 –> 01:20:21.899
Mark Roberge: Historically, it was better off finding product-market fit first, putting all the human resources there, and if you did it, then move to go-to-market fit on the market you chose.

518
01:20:22.130 –> 01:20:25.840
Mark Roberge: Because a lot of this can be automated, those could be done in parallel.

519
01:20:26.290 –> 01:20:31.700
Mark Roberge: Alright, so there are some nuances, and for the most part, it can be summarized as

520
01:20:31.860 –> 01:20:40.869
Mark Roberge: The learning cycles are significantly shorter, and some sequences that were… had to be sequential can be now done in parallel.

521
01:20:40.970 –> 01:20:42.179
Mark Roberge: If that makes sense.

522
01:20:43.010 –> 01:20:49.960
Julia Nimchinski: Absolutely. Mark, I’m curious on the Stage to Capital portfolio side of things.

523
01:20:50.130 –> 01:20:59.559
Julia Nimchinski: What mistakes are you seeing most common now with, you know, the new wave of founders and building agenda companies?

524
01:21:01.140 –> 01:21:06.840
Mark Roberge: they’re not leaning into it as much as they should. I think they’re being a little too, like…

525
01:21:07.390 –> 01:21:12.449
Mark Roberge: Need for, like, a quick win?

526
01:21:12.630 –> 01:21:17.110
Mark Roberge: Like, oh, we tried it, didn’t work. Well, what did you do? Oh, we, like, tried it for a day.

527
01:21:17.750 –> 01:21:24.279
Mark Roberge: You gotta, you gotta, like, it’s such a big opportunity for startups that it needs more…

528
01:21:24.690 –> 01:21:41.449
Mark Roberge: commitment. And the reason why I say that is, like, you asked a related question, is, like, what are some of the biggest innovator dilemmas out there that will cause today’s big company incumbents to be replaced by the native AI attackers?

529
01:21:41.630 –> 01:21:47.030
Mark Roberge: And I have a list of them, but the number one one I have.

530
01:21:47.640 –> 01:21:54.359
Mark Roberge: is that not how their product works, but how their organization works, and how their organization can be AI-first.

531
01:21:54.640 –> 01:22:03.139
Mark Roberge: You know, we’re starting to see that what used to take 10 engineers can now be done with 1. We will, in the future, see

532
01:22:03.740 –> 01:22:09.420
Mark Roberge: You know, reps that used to close a million in revenue will close 10 million a year.

533
01:22:09.750 –> 01:22:14.870
Mark Roberge: We’ll see finance books on the quarter be closed not in a month, but in an hour.

534
01:22:15.150 –> 01:22:18.170
Mark Roberge: We’ll see. I mean, it’s like, the list goes on.

535
01:22:18.440 –> 01:22:31.780
Mark Roberge: And it’s gonna be way easier for a ground-up native AI startup to exploit those performance efficiencies faster than a 10,000-person company that has to go through the change management process.

536
01:22:31.950 –> 01:22:40.920
Mark Roberge: And because of that, they’ll be able to pass those savings along to the customer, or deliver on higher, you know, profit margins. There’s a choice there that they have to make.

537
01:22:41.160 –> 01:22:47.270
Mark Roberge: And that will cause massive disruption across all categories. So, like, this…

538
01:22:47.880 –> 01:23:04.289
Mark Roberge: that’s the biggest opportunity. And, you know, it gets back to your first question, like, aren’t we already seeing that? Aren’t we seeing three-person companies that are getting to 30 million? It’s like, I just think that is, like, a combination of just good PLG and a little bit of, like, this hype flywheel.

539
01:23:04.290 –> 01:23:10.900
Mark Roberge: and experimentation revenue. It’ll come. And there are some examples out there where it’s real.

540
01:23:10.940 –> 01:23:22.579
Mark Roberge: But that’s not the majority. And I just think that founders need to make being an AI-run organization a bigger priority in order to disrupt the industry.

541
01:23:23.510 –> 01:23:36.030
Mark Roberge: Like, an extreme example, Julie, and this isn’t a good idea, but just to exploit it, kinda can’t wait for a rep just to be like, I’m just gonna rebuild Slack with 3 people, and sell it for 90% off.

542
01:23:39.310 –> 01:23:44.700
Mark Roberge: Like, it’s stupid, that’s not good, but you get the point. I’m trying to make an exaggerated comment, you know?

543
01:23:46.150 –> 01:23:47.719
Julia Nimchinski: What is defensible landmark?

544
01:23:49.170 –> 01:23:53.919
Mark Roberge: Yeah, it’s a great question. I think point of work is gonna become the biggest,

545
01:23:54.040 –> 01:24:00.430
Mark Roberge: system of action or point of work, people are saying it a different way. So, like, in, in…

546
01:24:01.310 –> 01:24:20.229
Mark Roberge: the last, you know, in both the on-prem and the cloud era, system of record was an extraordinarily powerful, bear, you know, moat or barrier to entry. Create huge switching costs. If you had the data in the database, and if you owned, like, where people would put it in.

547
01:24:20.400 –> 01:24:21.910
Mark Roberge: Then you are good.

548
01:24:22.360 –> 01:24:26.049
Mark Roberge: But I think that’s gonna get replaced by the point of work.

549
01:24:26.700 –> 01:24:33.149
Mark Roberge: Because if you are… if you own the point of work, you own the raw data, and then you can put it wherever you want.

550
01:24:33.350 –> 01:24:36.650
Mark Roberge: So, like, in our go-to-market arena, it’s like.

551
01:24:36.770 –> 01:24:53.070
Mark Roberge: it’s more important to own the transcript of the conversation between the rep and the buyer than it is to own the data of the CRM, in my opinion. And that would translate to product development, to, like, on the finance side, to own, like, where the contracts are signed and the invoicing occurs.

552
01:24:53.070 –> 01:25:12.809
Mark Roberge: That would be one example. I think the other example I’d throw out is, I do think pricing’s gonna be a nice innovation. I think there’s an economies of scale one around, having the most data to train the best models. Probably the biggest one, too, I like in the application layer is the place that I have the most

553
01:25:12.890 –> 01:25:22.970
Mark Roberge: conviction on as to where today’s businesses will… today’s startups will actually sustain is, vertical AI.

554
01:25:23.060 –> 01:25:31.960
Mark Roberge: Outside of the tech sector. I just think, like, today’s CTOs are… I think they’re trying to kill the application layer.

555
01:25:32.160 –> 01:25:44.960
Mark Roberge: I think they’re trying to, like, put all their data into, like, a single data lake, and then use the LLMs or horizontal tech, to, like, kill the functional divides between marketing, sales, finance, product, and engineering.

556
01:25:45.090 –> 01:26:00.709
Mark Roberge: And I think that… but I don’t think, like, the CTOs of, like, hospitals, banks, manufacturing plants have a build mentality. I think they just care about a vendor that has proven that they can go through the compliance demands of their vertical.

557
01:26:00.890 –> 01:26:08.149
Mark Roberge: And so I think, like, that’s where I’m most bullish on, is vertical AI companies, in non-tech sectors.

558
01:26:08.890 –> 01:26:11.300
Mark Roberge: And that’s a little bit of, like, a brand moat.

559
01:26:11.730 –> 01:26:14.529
Mark Roberge: Like, if you look at Harvey and legal, I think, like, I don’t…

560
01:26:14.690 –> 01:26:24.500
Mark Roberge: when a big law firm finds out that their peer is actually using AI, I don’t think they’re like, go do a vendor assessment and find the cheapest one. I think they’re like, figure out who they used and buy them.

561
01:26:24.850 –> 01:26:28.829
Mark Roberge: Right? So… so I think there’s a little bit of, like, a brand reputation out there.

562
01:26:29.710 –> 01:26:31.359
Julia Nimchinski: You only have 2 minutes, unfortunately.

563
01:26:31.360 –> 01:26:31.810
Mark Roberge: Don’t worry.

564
01:26:31.810 –> 01:26:35.419
Julia Nimchinski: As probably the last question.

565
01:26:35.900 –> 01:26:37.150
Deepinder Singh Dhingra: Sorry, I’m early.

566
01:26:37.470 –> 01:26:39.490
Julia Nimchinski: Yeah, one minute, Depender.

567
01:26:39.490 –> 01:26:40.050
Mark Roberge: Hey, dude, been here.

568
01:26:40.050 –> 01:26:40.620
Julia Nimchinski: brilliant.

569
01:26:41.230 –> 01:26:41.709
Deepinder Singh Dhingra: Yeah, the…

570
01:26:41.710 –> 01:26:47.320
Julia Nimchinski: So… The last question here is about B2B as an industry overall.

571
01:26:47.320 –> 01:26:47.900
Mark Roberge: Yeah.

572
01:26:48.260 –> 01:26:57.040
Julia Nimchinski: I’m really curious to hear your thoughts, since we… I mean, we do have agents in the equation, and we will transition at some point when they will.

573
01:26:57.040 –> 01:26:57.510
Mark Roberge: Yes.

574
01:26:57.510 –> 01:26:59.749
Julia Nimchinski: filter for buying, and…

575
01:27:00.130 –> 01:27:03.440
Julia Nimchinski: And the like. So, how do you see the industry transforming?

576
01:27:03.440 –> 01:27:21.769
Mark Roberge: Yeah, I think it’s just gonna be, like, the trust and relationship, to your point, goes away, Julia. I think we’re moving away from that. I think in the 80s and 90s, the best sales team won. Whoever had the golf membership and took the buyer to the golf and steak dinners, they won the account, and all the products sucked. I mean, shelfware, people hated it. It’s just the best…

577
01:27:21.850 –> 01:27:40.030
Mark Roberge: And, you were stuck with it. When we moved to the cloud, we’ve moved a little away from that. Like, you could find stuff out on your own, PLG. We still have, like, people buy from humans who you trust, blah blah blah, I want to make sure I can call them up and get stuff. I think that will completely go away. When buyer AI agents buy from seller AI agents.

578
01:27:40.030 –> 01:27:54.259
Mark Roberge: They’ll just make really good decisions that aren’t based on relationships. They’ll be experts at the needs of their business, and they’ll match it with the right vendor, and the seller AI agents won’t lie or manipulate. They’ll just be like, hey, they’ll just be honest, like, we’re not the right solution, you should go buy them.

579
01:27:54.260 –> 01:27:58.350
Mark Roberge: And so, like, that will change the paradigm quite a bit.

580
01:27:58.350 –> 01:28:13.959
Mark Roberge: But there’s, I think, a progression there. And if I could, Julia, just, like, I haven’t talked about the Science of Scaling book yet. I think it just went out for pre-order, like, last week. If you want to check it out on Amazon, you can. There’s another one named the same thing from the summer by a professor in the Midwest or something, so just be careful.

581
01:28:13.960 –> 01:28:38.830
Mark Roberge: you get the right one. I’ll do more of a massive launch in Q4, but I do want to say that 100% of the proceeds go to McLean Hospital, who is the globe’s leading research in mental health, so I just want to point that out as well. My last book went to a similar organization, and I think we’re in the midst of another, you know, big mental health situation, so if you do end up supporting the book, you are supporting that cause as well.

582
01:28:38.830 –> 01:28:39.650
Mark Roberge: when I say that.

583
01:28:40.400 –> 01:28:46.570
Julia Nimchinski: pitch pleasure mark, and never enough time. And yeah, we will be sending your book, actually, to everybody who’s watching.

584
01:28:46.570 –> 01:28:47.020
Mark Roberge: Oh, great.

585
01:28:47.020 –> 01:28:48.389
Julia Nimchinski: this figure, so yeah.

586
01:28:48.390 –> 01:28:48.770
Mark Roberge: Thank you.

587
01:28:48.770 –> 01:28:49.370
Julia Nimchinski: Have a thought.

588
01:28:49.540 –> 01:28:51.049
Mark Roberge: Alright, thanks so much, Julia.

589
01:28:51.050 –> 01:28:51.540
Julia Nimchinski: Thanks.

590
01:28:51.540 –> 01:28:53.060
Mark Roberge: Thanks, everyone. Wait.