This article is based on my conversation with Vadim Rogovskiy, founder of Rebel Capital and Eve, on the RolyPoly podcast.
Most founders I talk to are obsessed with staying in their lane. Vadim Rogovskiy's lane is finding a new lane. He built and sold an ad tech company, scaled a mobile monetization platform to $13M in revenue before it collapsed under him, co-founded a computer vision company in 2016 that's still profitable a decade later, and ran a VC fund. Two years ago he did what he calls a "factory reset" and built Rebel Capital — a holding company acquiring legacy software and service businesses and rebuilding them with AI.
What makes Vadim worth listening to right now isn't the résumé. It's that he's one of the few operators actually running the experiment most of us are still theorising about: what happens when you stop buying companies to consolidate them, and start buying them to rebuild them with agentic AI from day one?
The answer, based on one year inside his first acquisition, is that the post-merger integration playbook most PE shops are running is already obsolete. Not because the principles are wrong — cost-outs, offshore shared services, go-to-market upgrades — but because the biggest lever in an acquired SMB isn't hiding in the operating model. It's sitting in the founder's inbox.
Buy, Don't Build: The Entrepreneurship-Through-Acquisition Thesis
Vadim's shift into acquisitions came from a specific frustration with venture capital. After closing out his first fund, Geek Ventures, he realised that passive investing bored him — and more importantly, that he was leaving value on the table.
"Just cash as an investment is just a commodity," he told me. "I can give much more than just cash."
That's the core of his pitch for entrepreneurship through acquisition (ETA): instead of starting from zero and fighting for product-market fit, you buy a business that already has customers, revenue, reputation, and inbound leads. Then you apply AI and modern go-to-market on top. You get venture-style upside with private equity downside protection, because it's genuinely hard to kill a business that already has a book of customers paying real money. It's the same logic underneath any well-run buy-and-build integration strategy — the acquisition isn't the bet, the operating model on top is.
The subtler point — and the one I think most first-time acquirers miss — is the portfolio mindset. Vadim was explicit about this: the reason he left operating to build a holdco is that being one founder attached to one company for ten years is a losing game statistically. "If you play this game and you run one company at a time and it is a startup, the odds are not very high," he said. Shift to a portfolio of acquired businesses and the math changes entirely.
If you're thinking about roll-ups or buy-and-build, this reframe matters. You're not trying to hit a home run with one acquisition. You're building a portfolio where no single business has to be exceptional for the whole thing to compound.
The Hidden Revenue Sitting in the Founder's Inbox
The most concrete moment in our conversation — and the one I've been thinking about since — was Vadim's description of what happened the day Rebel Capital acquired their first company.
The business is in event infrastructure: running trade shows and business conferences for other companies. The founder had been running it for 18 years and wanted to retire. Pretty typical search fund territory. What wasn't typical was what Vadim found when he plugged the founder's Microsoft inbox into Eve, the AI tool his team had been building in parallel.
Roughly 400 leads, completely untouched for the last six years.
Not cold leads. Not junk. Genuine inbound prospects that had emailed the founder and slipped through the cracks because there was no CRM discipline, no follow-up system, and one person trying to run sales on top of everything else. Vadim's team handed that list to the in-house salesperson, who closed several six-figure deals off it within three to four months.
This isn't a one-off. Vadim described doing the same exercise with a commercial real estate company and finding $20 million in hidden pipeline sitting across spreadsheets, mailboxes, and fragmented tools. The data was always there. Nobody had ever connected it.
Here's why this matters for anyone thinking about post-merger integration: when you buy an SMB, you inherit a revenue function that almost certainly looks like this. Founder-led sales, CRM half-used, proposals in email, contracts in Dropbox, transcripts in some notes app nobody opens. The integration playbook says "consolidate the tech stack" — and a proper CRM migration after acquisition or data migration is genuinely necessary work. But consolidation alone doesn't surface the value — it just tidies up what's already visible. The AI layer is what finds what's invisible.
This is the part of Vadim's operating model I think most serial acquirers will copy within the next 24 months. Running a pass through inherited sales data in week one of an acquisition isn't a technology project — it's revenue hygiene, and it pays for itself almost immediately.
Why You're Probably Holding Too Long
We spent part of the conversation on exits, and Vadim was refreshingly blunt about the mistake that cost him tens of millions of dollars at his second company, Flicky.
Flicky was a mobile monetization platform doing $13M in annual revenue. Vadim had a buyer lined up at a $12M valuation — a steep discount from what he thought the company was worth, but a real offer. He kept pushing, hoping for more. Then the ad tech market cratered from fraud and fake traffic, the deal fell apart, and he eventually had to close the business entirely.
His takeaway, which he'd posted about on LinkedIn days before our call: "I've met so many founders who regretted waiting too long to sell. I've never met a single founder who regretted selling too early."
The reason, he thinks, is structural. Founders are by definition delusional optimists — that's what lets you start a company in the first place. But that same optimism distorts exit timing. You always believe next year's number will be better. Sometimes it is. Often the macro moves against you, and the window closes.
I think about this a lot in the context of serial acquirers talking to founders. The founder staring at your LOI is running exactly this calculation, and most of them are going to lean toward "one more year." The acquirers who close deals consistently are the ones who make it easy for a founder to see the case for selling now — not by pressuring, but by surfacing the opportunity cost of waiting. Vadim's own experience is useful ammunition: even when your business is growing, the exit window is almost always narrower than you think.
The Vertical AI Moat: Why Horizontal Tools Will Get Flattened
Vadim spent a decade in AI before it was fashionable. His view of where the value accrues in this next wave is worth taking seriously.
His thesis, in short: horizontal AI tools are going to get demolished by the foundation model companies themselves. OpenAI, Anthropic, and the rest are incentivised to build picks and shovels — the generic capabilities that power everything. Anything you build that's "a thin wrapper" on top of those capabilities is at permanent risk of being absorbed into the base platform next release cycle.
The defensible play is going deep into a specific vertical. Harvey in legal. Vertical SaaS selling into hospitals or investment banks. Tools that integrate with messy, proprietary, hard-to-reach workflows and collect domain-specific data nobody else can access. "If it's easy, I don't do it," he told me. "You work on hard things. Integrating tools that don't talk to each other — it's very hard. If you do it, it's going to be hard to displace you."
This is essentially the same argument I'd make for PMI work specifically. The reason IT post-merger integration doesn't get automated away by a generic AI tool is that the messiness is the moat. Every acquired company has a different finance system, a different CRM, a different set of half-documented workflows, and a different culture around data. An AI assistant that knows how to handle that — the ugly reality of a mid-market consolidation — is worth more than a thousand horizontal copilots. This mirrors what we see at PMI Stack, where the mapped 30+ AI application points in our integration process are all about navigating the specific messiness of systems migrations, not generic knowledge work.
The Roll-Up Playbook, Rewritten for Agentic AI
When I asked Vadim about how agentic AI changes roll-up strategy, his answer surprised me — not because the points were novel, but because of how matter-of-fact he was that the playbook has already shifted.
At the operating company level, he's using AI for two main loops. The first is revenue: Eve connects to inboxes, CRMs, spreadsheets, and meeting transcripts, builds a single source of truth, and deploys an agent to reactivate dormant leads. The second is delivery: a project management agent that coordinates service delivery across multiple parties. At his event infrastructure company, this combination allowed them to shrink the operations team from four people to one — while the business continued running events for clients.
At the holdco level, he's getting cross-company revenue visibility for the first time. A portfolio operator running a roll-up historically had to wait for monthly board packs to see what was happening in pipeline across five, ten, fifteen companies — which is exactly the pain that traditional post-acquisition reporting consolidation projects exist to solve. With an AI layer sitting across all of them, that visibility becomes real-time. The implication for integration is significant: you no longer need to fully merge systems to get the management information a consolidated group needs. A layer on top can do what used to require a multi-year ERP consolidation.
Vadim was also candid that this is changing how he thinks about the integration sequence in his own roll-up. "Now we look at it differently, because integration of companies should happen much faster and requiring fewer resources," he said. That's a significant statement from someone actively running acquisitions. If he's right — and the economics of his first deal suggest he is — then PE sponsors budgeting 18-month integration timelines are budgeting for a world that no longer exists.
One at a Time: The Integration Sequencing Debate
One thing Vadim was careful about was the pace of integration inside a roll-up. His team is currently looking for a second, larger "platform" acquisition in event infrastructure, with Aleterra — their first acquisition — likely folded into it rather than the other way around. After that, the plan is to add companies one at a time rather than merging two or three simultaneously.
"We thought, okay, we're going to acquire two companies and merge them with [the first one], but it raises the complexity, raises the risk," he said. "It sounds great because you can faster get to a platform, but it just raises complexity so much. The approach is to go one by one until we get more experience."
This echoes something Sam Turner said on an earlier RolyPoly episode: early on, keeping businesses separate feels less risky, but after a certain number of acquisitions, not integrating becomes the bigger risk because you lose standardisation and can't deploy shared capabilities (AI included). It's the same reason mature acquirers eventually build a serial acquisition integration playbook — once you're doing three to five deals a year, ad-hoc integration stops working. The tipping point depends on the operator's own bandwidth and the maturity of their integration playbook. Vadim's position — integrate deliberately, one at a time, until the muscle is built — is the honest answer most first-time roll-up operators need to hear. Announcing "Platform 2.0" with four simultaneous integrations in year one is how good deals turn into distressed ones.
What Operators Should Actually Take From This
If I compress Vadim's operating model into a few things a serial acquirer could run with this quarter, it would be this:
- Run an AI pass through inherited sales data in week one. Before you touch the operating model, before you restructure the team, connect inboxes, CRMs, and spreadsheets to a single system and look for dormant pipeline. The ROI on this is almost always faster than anything else in the integration plan.
- Build the portfolio, not the identity. If you're still attached to one company's success as your success, you're not running a holdco — you're running a startup with extra steps. The shift is mental before it's operational.
- Bias toward exits, not extensions. If you're advising founders in your network, share Vadim's reframe: nobody regrets selling early. The sellers who time it well are the ones who treat exits as portfolio moves, not identity decisions.
- Go vertical on AI, not horizontal. The AI investments that pay off inside a roll-up are the ones that know your specific industry's workflows. Don't buy a generic copilot and hope; build or buy something that speaks your sector's language.
- Integrate sequentially until you have the muscle. The temptation to batch acquisitions and merge three at once is real. The learning curve is steeper than it looks. One at a time, then accelerate.
Roll-ups aren't a new asset class. But the version of the roll-up playbook being written right now — AI-native, portfolio-minded, integration-light at the infrastructure layer and heavy at the revenue layer — is genuinely new. Vadim's one of the people writing it in real time. Worth listening.
