[Un]Churned · 30 Sep 2026 · From the week of 28 September
Can AI Really Manage Your Entire Renewal Book? ft. Grant Clarke (Gainsight))
These are notes on the conversation, checked against its transcript. The episode itself has the full discussion.
In brief
Host Josh Schachter interviews Grant Clarke, EVP & GM of Gainsight Atlas and a veteran of about 15 years at renewals outsourcer ServiceSource and later Dropbox. Atlas is an 'AI Native Services' (AINS) business that manages a company's renewal book, starting with the long tail, by pairing AI agents with human renewal managers, and customers pay for the outcome rather than for software. Grant covers how Atlas onboards customers, why data hygiene and willing partners matter, how agents triage and hand work to humans, and how a loop with forward deployed engineers turns repeated human escalations into new agent capabilities within weeks. His central argument is about economics. Disciplined customer selection early on, then shifting the share of renewal actions handled by agents from about 20% to 55%, is how he says AINS avoids the margin compression of traditional BPO and heads toward the roughly 70% gross margin benchmark.
For founders
- Grant Clarke argues that AI-native services avoid the classic services margin trap by not taking every customer early: pick design-partner customers who build with you, accept a flatter revenue curve, productize, and then scale. Josh summarized this as 'go fast, go slow, go fast again.'
- Grant frames AINS margin as a unit-of-work problem: count every action in the delivery process, then raise the share the agent handles (Atlas targets moving from 20% to 55% agent-led actions) so each new customer needs fewer humans (maybe six or four instead of 10).
- Josh Schachter cites Jake Saper's gold standard of 70% gross margin for AINS companies, against roughly 80% for SaaS, and notes that early-stage AINS startups have not yet reached it.
- Grant describes a fast learning loop in which forward deployed engineers, PMs and business analysts sit beside the human operators, classify recurring escalations, and ship agent capabilities in roughly two-week sprints.
- Grant says Atlas is paid for the renewal outcome rather than for software, and that it differs from BPO by putting AI agents in the loop with human renewal managers. Owning the outcome means finding the leverage ratio that keeps delivering it, not cutting staff just to hit margin.
For revenue leaders
- Grant Clarke estimates each renewal involves 10–20 actions, about 10 for a customer seeing value and 20–25 for an angry one; multiplying actions by contracts gives a 'renewal action inventory' to split between agents and humans.
- Grant assigns agents early-stage outreach and value-based nudges, such as pointing out unused features alongside training content, and expects humans to handle more of the late-stage negotiation, angry customers and odd edge cases.
- Grant's Atlas cockpit has the agent research accounts overnight, handle what its guardrails allow, and give the rep a pre-prioritized queue with escalations such as at-risk customers at the top.
- Grant says long-tail renewal coverage remains the most painful problem, with the same conversation in 2026 as in 2006, and that AI agents plus humans now make it manageable.
- Grant says outsourcing a renewal process without senior sponsorship to change it just preserves a deficient process. He frames onboarding as a diagnostic of coverage gaps, with co-investment toward a shared outcome such as, hypothetically, improving GRR by three or five points.
What was said 20, most useful first
Gainsight Atlas charges for a managed renewal outcome rather than for software, combining AI agents and human renewal managers to run a client's renewal book. Listen
Grant Clarke describes Atlas as an operational framework that combines an AI agent with a renewal rep to manage a renewal book on the client's behalf. The client pays for the outcome, not for the product or the software. Josh Schachter summarizes it as 'give us the keys,' with Atlas acting as the client's renewal managers and Grant calling it an extension of the client's team.
“It does that in a way that we are delivering an outcome so the company pays for the outcome not for the product not for the software.”
Atlas turns recurring human-handled escalations into agent capabilities through forward deployed engineers who ship them in roughly two-week sprints. Listen
Grant Clarke says Atlas records what reps do outside the agent's handoffs and classifies those actions into buckets. Forward deployed engineers, product managers and business analysts sit 'in the boat' with renewal managers to monitor this. When a category of escalation recurs multiple times, it becomes a product requirement. The FDE team develops it in maybe two weeks, tests it for a few weeks, and puts it back into Atlas.
“The FDE team can sprint, develop that in maybe two weeks, run a few weeks of testing, and then it's right back into Atlas to handle the next customer issue that comes up.”
Grant Clarke applies Emergence Capital's 'Mirage PMF' idea to AINS: early inbound interest should be rationed, accepting slower revenue while you productize, before opening the floodgates. Listen
Grant credits Jake Saper's AINS playbook at Emergence Capital. In that playbook, early interest from many customers is a 'Mirage PMF', and taking them all on risks diluting the autonomous leverage you are building. The approach is to choose customers who participate in the build, accept a flatter early revenue curve, and focus on learning loops and proof points that the trained model succeeds more and fails less. Once the knowledge base, guardrails and evals are productized, you can take on 10 more customers at once without proportional cost. Josh Schachter calls it 'go fast to go slow to then go fast again,' and both add that it takes discipline.
“And you may have a little bit of a flattening of the curve early on in revenue. but you're doing that because you're working on the signals.”
Josh Schachter cites Jake Saper's 70% gross margin as the gold standard for AI Native Services companies, compared with about 80% for SaaS. Listen
Josh Schachter says SaaS gross margins are around 80%, with 80-plus considered very good, and that Jake Saper believes AINS companies can reach 70%. He adds that the AINS startups Jake works with are early in their maturity and lack the data and reps to have reached that level yet. Grant Clarke says he believes Atlas can get there.
“Jake talks about the gold standard being 70% of margin to get there. You know, in SaaS, it's around 80%.”
Raising agent-led renewal actions from 20% to 55% is the source of margin accretion, letting Atlas staff maybe six or four humans instead of 10 for the same contracts. Listen
Grant says this is what separates Atlas from a BPO that simply deploys people to do renewals. If Atlas-led actions go from 20% to 55%, then each new customer needs maybe six humans plus Atlas, or four, rather than 10 humans plus Atlas for the same number of contracts. Productizing that action-level autonomous leverage is how he expects to keep deploying economically into new customers without margin compression over time.
“We're not deploying the 10 humans plus Atlas. We're maybe deploying six humans plus Atlas or four for the same number of contracts.”
Grant Clarke separates AI Native Services from traditional BPO by having an AI agent work in the loop with humans, where BPO relies mainly on people and process with a little tech. Listen
Grant acknowledges that Atlas can sound like outsourcing or BPO, having worked at the outsourcer ServiceSource for 15 years. He says BPO 'has a lot to do with the people in process and maybe a little bit of tech.' The differentiator he claims is putting the AI agent in the forefront of the renewal process alongside the human, which he says delivers the scale companies have wanted for reaching the long tail.
“What we're really trying to differentiate here is we're bringing AI agent in the forefront as part of being in the loop of that renewal process with the human.”
Deploying AI agents in renewals is hard, and Atlas absorbs that burden by deploying humans alongside the agent so it can be trained faster. Listen
Grant says many companies are struggling to deploy AI agents across technical services, early funnel, sales and renewals, and that it is 'difficult enough just to get the process to work with a human.' Atlas aims to build the frameworks, standards and loop process to train the agent faster because humans are deployed with it. It offers this as a service so the client company does not carry the burden.
“It's difficult enough just to get the process to work with a human.”
Every new workflow Atlas learns gets trained back into the system, with domain experts' years of experience serving as the initial 'down payment.' Listen
Grant describes 20 years in the 'engine room' of renewals and says you can now teach an agent model based on that hard-won experience. Each time the team finds a new workflow or a new way to approach a problem, it is trained back in so the next customer gets started faster. The initial investment is his experience and that of other decade-plus renewal veterans he plans to hire.
“every time we're going to come up with a new workflow, understand a new way to come at the problem, we're just training that back into what is going to help the next customer get started faster.”
Atlas begins each engagement with a diagnostic of how the customer currently covers accounts, then proposes co-investment to close the gaps, framed around improving GRR by a few points. Listen
Grant Clarke says prospects vary: some have at-scale digital or auto-renewal motions, some already use a BPO, and some have CSMs who triage down into the lower tier. Atlas runs a straightforward diagnostic of coverage elements against the benchmarks it needs to launch. It then comes back as a partner to say where both sides can co-invest to make an agentic motion work. He frames the shared goal with a hypothetical GRR improvement of three or five points, and says iteration continues through implementation and steady state.
“If it means that we can improve GRR by three points or five points, wouldn't that be interesting?”
Listen to the episode Retention & customer success Link to this
Everyone's renewal data is rough, and Atlas handles this by defining what 'renewal-ready data' looks like and resolving gaps jointly rather than treating them as a showstopper. Listen
Grant cites constant changes of Salesforce instances and CRMs, and multiple conflicting versions of what a customer owns across systems. Atlas compares the customer's data with its own definition of renewal-ready data and decides with the customer how to close the gaps. Clean data matters both for the agent's value-based engagement with customers and for giving the human rep the right context.
“What I would say is we understand what renewal ready data looks like.”
Listen to the episode Retention & customer success Link to this
Atlas's first target customers are existing Gainsight CS customers, because the agent can draw on their health scores, CTAs and playbooks. Listen
Grant Clarke says the first interest is Gainsight's own install base, especially customers who have embraced the customer success methodology and platform. Atlas can then layer on everything Gainsight CS provides, including customer access, health scores, CTAs and playbooks. He calls this a huge advantage for the first set of customers.
“It's a huge advantage because we can layer and tap into everything that Gainsight CS offers as part of access to customers, health scores, CTAs, playbooks, all those things.”
Clients who 'throw renewals over the fence' BPO-style won't get Atlas's full value; the right fit has senior sponsors willing to change processes across groups. Listen
Grant wants partners who build with Atlas rather than outsource in the classic way. He looks for senior leadership that wants to rethink customer coverage and is willing to change processes across different groups. Without that, he says, you are just running a process that stays as deficient as it started. He calls this partnership profile especially important early on.
“Otherwise, if you throw something over, you're running a process on something that's already going to stay deficient from where it started.”
In the Atlas 'cockpit,' the agent works the renewal book overnight, handles what its guardrails allow, and hands the rep a pre-prioritized queue led by escalations. Listen
Grant Clarke says renewal reps normally start each day behind and triaging inboxes and CRM tasks. In Atlas, the agent researches overnight or in the background, checking each account's last touch and any responses to outreach. It then decides whether it can act autonomously under its guardrails, rules and deterministic settings, or must escalate or hand off to the human. Escalations go to the top of the list, such as a customer saying they have three technical issues and won't renew, and the rep picks them up with context from the agent.
“And what the agent is trying to decide is, can I as the is the autonomous agent handle this based on my guardrails, my rules, the deterministic settings that I have.”
Services businesses compress their margins by saying yes to every expansion request before working out how to scale cost below the pace of revenue. Listen
Grant says the trap for legacy service providers is that when a customer offers more work, such as more customers or new regions, the quick answer is yes because it drives top line. The provider then never finds time to work out how to scale cost over time, and costs rise with revenue. Automation, RPA and process improvements help, but he says they yield only smaller degrees of improvement.
“the very quick answer is yes, we'll figure it out because that's driving top line, but then you never have time to really figure out when you scale, when you put the cost in place, how do you continue to scale it over time?”
Grant Clarke estimates each renewal involves 10 to 20 actions, from about 10 for a satisfied customer to 20–25 for an angry one, and sums them into a 'renewal action inventory.' Listen
Grant treats the renewal action as the real unit of work that drives cost. The count expands or contracts with how the customer engages: a really angry customer may take 20 or 25 actions, and one seeing value closer to 10. Adding up actions from early to late stage and multiplying across all contracts, for example for a set of 5,000 customers, gives the total renewal action inventory that Atlas works to automate.
“So really angry customer, they're probably like 20 or 25. The customer that's seeing value, they understand what they're batting at, probably closer to 10.”
Listen to the episode Retention & customer success Link to this
Grant Clarke assigns renewal outreach and value-based nudges to the autonomous agent, and expects humans to handle more of the late-stage negotiation, angry customers and unusual scenarios. Listen
Grant gives outreach-and-response as an example of work suited to the agent's back-and-forth. Another is a value-based message noting the renewal is coming up, pointing out a feature the customer isn't using, and linking a white paper or training module. He expects humans to do more of the late-stage negotiation, the 'pissed off' customers and odd scenarios. He says the agent can still work through more of the action inventory over time.
“And then when you get into the late stage like negotiation or the pissed off customer or the one that's got some random scenario to play out, we know that the human's going to do more of that.”
Listen to the episode Retention & customer success Link to this
Because Atlas owns the outcome, it sets the human-to-agent leverage ratio by what keeps the outcome delivered, not by cutting staff to reach a margin target. Listen
Grant expects fewer reps assigned per account over time, but says that has to be tested and proven against the outcome. Since Atlas owns the outcome, it is responsible for finding the right leverage ratio that keeps delivering it. He says Atlas won't 'scrape it just to get margin.'
“We're not going to try to scrape it just to get margin. We've got to deliver the outcome in order to achieve that.”
Grant Clarke expects Atlas to expand from long-tail renewals into pre-renewal nurturing and customer success onboarding, which he calls a more repeatable process suited to guardrails and rules. Listen
Grant says Atlas is starting with long-tail renewals because that is where customers feel the most pain. He sees possible expansion into nurturing ahead of the renewal and into customer success onboarding (not technical onboarding) to increase adoption. He calls onboarding more repeatable, so more guardrails and rules can be built for it. He also raises the question of how this could look in the sales cycle.
“It could be nurturing ahead of the renewal. It could be going into maybe even onboarding process, which is more of a repeatable process anyway that you could drive more guardrails and rules for.”
Listen to the episode Retention & customer success Link to this
An agentic judgment layer will soon help renewal reps handle subjective edge cases, though he agrees it may enable reps rather than act alone. Listen
Grant says renewals constantly surface obscure objections. His example is a customer saying a since-departed sales rep promised something in writing and threatening to move to a competitor. He believes that as Atlas accumulates context from more of these cases, an agentic layer will sift through it and give reps a much better way to handle them. When Josh Schachter suggests it may 'at least enable' reps, Grant agrees it may not do it by itself.
“I do believe the agentic layer is going to soon be able to really leap forward on context to handle every random nuance that may come up.”
The long-tail renewal coverage problem is unchanged in 20 years, and starting there then expanding coverage once value is proven worked at ServiceSource. Listen
Grant says the long tail is where companies feel the most pain, because it is hard to scale cost and engagement for many smaller customers while also retaining the large accounts that drive most ARR. At ServiceSource they started with the long tail. Once they proved value, they quickly found other ways to increase coverage for the client. He says the 2026 conversation is the same as the one in 2006.
“here we are in 2026, and I swear it's the exact same conversation from 2006.”
Listen to the episode Retention & customer success Link to this