Most SaaS companies talking about AI right now are talking about copilots, ways to make existing software look smarter without changing what they actually sell. Gainsight did something more drastic. It built an agent that can run a customer renewal end-to-end, tried to sell that agent as a product, watched the pitch fail, and made the harder call: stop selling the software and start selling the renewal itself. It's a live test of the SaaS-to-AI native services (AINS) transition most companies are still only talking about, and it's running inside a private-equity-owned business with real debt obligations and a large software franchise it can't afford to walk away from.
Gainsight was acquired by Vista Equity Partners in 2020 for a little over a billion dollars and today runs several hundred million dollars in SaaS revenue. Chuck Ganapathi, who joined as President & COO in 2024 and took over as CEO roughly a year ago, announced at the company's customer conference a few months back that Gainsight would start delivering retention as a service, selling outcomes directly. At our AINS event in New York this summer, General Partner Jake Saper sat down with Chuck to talk through the decision, the pushback, and what's happened since.
Jake: Why make this change to selling outcomes at all? You're running a large, profitable software business.
Chuck: It started in 2025 as the world was waking up to the power of agentic AI. We asked who's going to build the agentic version of a customer success manager. Not to replace the human CSM, but to work alongside them. We broke down everything a CSM does and landed on renewals as the clearest place to start. We built a renewal agent and took it to a few pilot customers.
It didn't go how we expected. When you sell someone software, they want to buy software. They ask about seat price the same way they always have. And our technology was moving faster than customers could adopt it. We call that the diffusion gap. Bringing in an agent-led workflow with their existing process asked for organizational change most teams weren't ready for. That's when we stopped trying to sell the tool and decided to own the whole outcome instead.
Jake: You're owned by a private equity firm that bought a software company. What was that conversation like?
Chuck: Vista saw the end of seat-based pricing earlier than most, so they were open to it. PE understands margin inversion, but the version they know is buying a low-margin business and cutting opex (R&D, sales and marketing, G&A) to make it high-margin. Historically, they bought software companies with great gross margins and very high CAC, then cut the CAC to expand margin. What I had to explain was that this version runs through cost-of-goods-sold instead.
I wrote an internal post for Vista's leadership and called it "COGS is the new CAC." In an agentic business it flips: CAC is low, and the cost lives in COGS, because the tokens and the humans in the loop both sit there. Cost starts high because agentic delivery is expensive up front, and the bet is that AI autonomy brings it down over time. Once I framed it in language they already used for evaluating other businesses, they got it immediately. It also helped that outcome pricing taps the labor budget, not the IT budget, and that’s a much bigger pool.
Jake: Who do you hire to build something like this inside a company that's never done it before?
Chuck: We built Atlas, our AI-native services business, as effectively a separate company inside Gainsight. The first hire was Prem Parameswaran, a former co-founder of mine who'd gone from AWS to WeWork to a technology-enabled HR outsourcing firm: someone with a track record of moving from pure engineering into technology-enabled services. Our core software business essentially funded his small team the way a venture firm would fund a startup.
Once we decided to build the actual services layer, we needed someone to run the operational side, blending human and agent work. We found Grant Clarke, twenty years in BPO before moving to Dropbox to run revenue operations and analytics. Deep services experience paired with real exposure to an AI-native engineering culture is rare, and it turned out to be exactly what the role needed.
Jake: Where does Atlas actually land inside a customer? You're not walking in and offering to take over their whole renewals team.
Chuck: No. We found our landing wedge via the Pareto principle: 80 percent of the revenue comes from 20 percent of the customers, the big enterprise accounts, and that's where all the attention goes. Then there's a long tail, thousands of small accounts that add up to the other 20 percent. It’s always been a hard segment to serve and they've never been worth the cost of a human team, in-house or outsourced. So companies ignore them and accept a high churn rate. You don't want to kill that revenue, but it’s hard to justify spending expensive human capital on it. We go to that company and say, give us that $50 million of ARR and we'll renew it for you. That becomes our revenue under management, the way a fund has assets under management. A customer paying you five grand a year has never gotten the white-glove treatment. Now agents, with humans in the loop, can treat them like your best customers, with economics that finally work.
Jake: I want to flag the landing strategy here, because it generalizes. There are two ways to enter a market as an AINS company. One is head-to-head with the incumbent provider: do the same job faster, better, and cheaper. The other is what you did: find the work nobody was doing because nobody could afford to do it, and take that. It's a version of the Jevons paradox. When a service gets cheap enough, people don't just switch vendors, they buy more of it than they ever did. You're not displacing anyone on these accounts. You're getting paid for renewals that used to be left to chance, and the expansion paths come from the same place.
So what's changed since you launched?
Chuck: First was pricing. We priced against revenue under management, but our own sales team kept defaulting back to flat contract pricing, because it was simpler to sell and easier to earn commission on. Selling outcomes looks more like consulting sales than software sales, and that took real retraining.
Second was what I'd call emergent work. Agents perform well enough that they create demand nobody planned for. A long-tail customer that used to get almost no attention starts getting real, responsive service, and suddenly that customer is asking for things we never scoped, and telling other customers about it. One started emailing our agent about everything because it was the fastest way to get an answer. We're still figuring out where to draw the line, because chasing all of it without discipline is exactly how the margin math falls apart.
Jake: How do you know if any of this is actually working?
Chuck: For our customers, the outcome that matters is GRR and NRR. What I watch internally is the Autonomy Rate. What share of the individual actions in a renewal (outreach, negotiation, paperwork) an agent completes without a human stepping in. And revenue per renewal manager, the human-in-the-loop equivalent of ARR per FTE. If that number isn’t climbing, we’re just scaling like a staffing firm. It's early, but we know the shape of the curve we're on.
Jake: Five years out, does the software business still exist?
Chuck: It does, and it stays healthy because we have thousands of customers who depend on it to serve their customers. We have debt service to make every year that a venture-backed startup doesn't, so I don't have the option to “burn the bridges” the way some companies like Intercom have done. But the software business alone can't be how we grow anymore. The market for customer success software has an upper ceiling. The real growth path is going deeper into our own space instead of chasing adjacent markets in which we don't have a right to win. My expectation is the AINS business has a TAM that is five times the size of the software business, while the software business keeps performing well in its own right at traditional PE-style growth.
Jake: What makes a company a good candidate for this kind of shift?
Chuck: Finding the goldilocks zone. The function that the customer hires you to do has to be important enough that they feel some pressure around it, but not so critical that they'd never hand it to anyone outside their own team. Repetitive enough to automate, but with enough real variation that it isn't just RPA. And the customer experience has to matter: high stakes, but not the highest stakes in the business.
For us, that's the long tail customer base of our clients: real revenue, real relationships, but not the enterprise accounts a client would never fully hand over. Find that zone in your own business, and I think this is where you find gold.
.jpg)
