Three CRM Stories, One Question: Who Pays for AI

Question: How are CRM vendors pricing AI agents right now, and is that changing?

Quick answer: CRM vendors haven't agreed on how to price AI agents. Some buyers are opting out entirely: 35% of enterprise teams have already built their own tools instead of paying for one. Sierra is testing pay-only-if-it-works pricing. But both bets skip over the same cost. Keeping a platform current takes dedicated, ongoing work, whoever's doing it.

Three unrelated-looking stories came out of CRM and enterprise software this week. A pricing report. A CNBC interview. A split decision on Wall Street. Read on their own, they're just news. Read together, they're the same story. Nobody in this industry, whether vendor, analyst or buyer, has actually settled on how AI agents should be priced. Until that gets settled, everyone downstream is guessing.

A third of enterprise teams have already built their own replacement

A widely cited 2026 Retool report found that 35% of enterprise teams have already replaced at least one SaaS product with something they built in-house, and 78% expect to build more this year. That lines up with what's happening at the SMB end too. Smaller companies are swapping out specific CRM subscriptions for AI-built alternatives as well.

That number moves this conversation out of anecdote territory. A year ago, "we built our own CRM instead of paying for one" was a case study you'd point to. Now it's a third of enterprise teams doing some version of it. The tool was never really the constraint. A capable person with the right AI tools can stand up something functional fairly quickly now.

What that stat doesn't answer is who's maintaining it in six months. Security patches, integration breakage when a connected tool changes its API, adapting the thing as the business changes: that work doesn't disappear because you built it yourself instead of buying it. I've been going through my own version of this inside Engagent this year, in small, deliberate steps, using Claude and Claude Code to build things we used to either pay for or do by hand. The pattern holds at our scale too. Building it is the easy part. Some share of that 35% is quietly under-resourcing the part that comes after.

Sierra says: don't pay us until the AI actually works

Quick context: who is Sierra? Sierra isn't a feature inside an existing CRM. It's its own independent company, founded in 2023 by Bret Taylor (Salesforce's former co-CEO) and Clay Bavor (an 18-year Google veteran). Venture-backed and now valued at $15.8 billion, Sierra builds AI agents that plug into a company's existing systems to handle customer service work directly. It competes with platforms like HubSpot and Salesforce for the AI-agent layer; it isn't part of any of them.

In a July 14 CNBC interview, Sierra co-founder Clay Bavor laid out the company's pricing model: customers pay only when an AI agent actually resolves something, whether that's closing a support ticket, saving a customer who intended to cancel, or completing an upsell. Not a flat fee per seat. Bavor said that model has pushed the company's proof-of-concept conversion above 90%, and argued AI agents broadly could push software away from seat-based licensing altogether.

Every major CRM vendor right now, HubSpot's Breeze, Salesforce's Agentforce, SugarAI's revenue intelligence layer, is still charging for AI the old way: bolted onto a per-seat licence. If outcome-based pricing catches on and buyers genuinely prefer it, that's not a small feature update. It changes the question a client should be asking a vendor: not "does this AI feature work," but "am I paying for the seat, or the result."

My take: I don't fully buy the "pay only when it works" framing. An agent doesn't show up finished. Someone still has to build and configure it for what they actually need it to do. If it's set up wrong, that's not automatically the vendor's failure. What I think a platform owes you is different: the right mechanisms to build and run the agent properly, and a real commitment to keep developing that toolset over time. Whether the agent executes well after that is mostly on how you set it up and use it. That's still worth a licence. You're paying for the infrastructure, the platform and the ongoing development, not purely for one outcome landing.

Wall Street can't agree on Salesforce's AI either

Guggenheim upgraded Salesforce this week, arguing the case that AI will gut demand for enterprise software is overstated. Shares moved up roughly 5% on the call. Within days, KeyBanc and Phillip Securities went the other direction, citing soft feedback on Agentforce sales. Meanwhile Salesforce's own numbers show Agentforce annual recurring revenue near $1.2 billion, with combined AI-and-data revenue approaching $3.4 billion, both growing at a solid clip year over year.

This isn't hype versus scepticism in the abstract. It's two sets of professional analysts, looking at the same real numbers, reaching opposite conclusions about whether that revenue is durable. If people whose full-time job is reading Salesforce's earnings can't agree, that's worth remembering the next time any vendor, Salesforce or otherwise, pitches an AI-agent roadmap with total confidence.

Where these three stories actually connect

Pull on the thread and the first two stories here are arguing the same point from opposite directions. The build-vs-buy stat shows a third of enterprise teams willing to build their own tools instead of paying for one. Sierra's pitch shows a vendor betting people would rather pay for results than a seat.

What connects them is this: in both cases, someone still has to do the unglamorous work of keeping the thing current: security patches, API changes, adapting to what people actually need next. That work doesn't disappear because you built the tool in-house, and it doesn't disappear because a vendor rebrands its pricing as "outcome-based" instead of per-seat.

The real question isn't whether AI makes it possible to build your own CRM. It clearly does. It's whether whoever's doing that ongoing upkeep, in-house or vendor, actually has people whose full-time job is staying dialled into it. In my experience, most organisations don't. I've seen businesses that don't even have a dedicated, up-to-date in-house marketing function; they're still leaning on outside agencies for that. If that's the reality for something as core as marketing, expecting the same organisation to sustain a self-built CRM's ongoing development is optimistic at best.

Same logic applies on the vendor side. If Salesforce's or HubSpot's own analysts can't agree on whether their AI revenue is durable, that's a live example of how unsettled this still is even for the companies whose full-time job this is.

Which is why I still land on this: pay for a platform whose provider treats this as their day job, not a side project, and expect that licence to cover a genuinely current, well-maintained tool, not just a login.

Closer to home

I want to bring this back to where I actually work. As far as I'm aware, there isn't yet a Caribbean-built CRM out on the open market, something homegrown that a business here could simply buy off the shelf the way you'd buy HubSpot or Salesforce. What I am seeing more of is organisations reaching out to local providers to build something custom instead.

That's encouraging, and it's also exactly the pattern this whole piece has been circling. A custom build still needs someone available locally who can configure it, support it and keep adding to it as the business changes. In a region where technical talent is harder to find and retain than in bigger markets, that ongoing capacity question is sharper here, not softer. It's part of why we've spent this past year building out our own regional delivery team at Engagent, rather than treating implementation as a one-and-done project.

I don't think this piece needs to be a Caribbean story to matter here. But if you're a Caribbean business weighing a custom CRM build against buying an established platform, the question from earlier in this piece is the one worth asking first: who's actually going to be around to maintain it, and is that genuinely their day job?

Key takeaways:

• 35% of enterprise teams have already replaced a SaaS tool with something built in-house; 78% plan to build more. Build-vs-buy is now a mainstream decision, not a fringe one.
• Sierra's outcome-based pricing model (pay only when the AI resolves something) is a direct challenge to the seat-based pricing every major CRM vendor still uses for AI features, though it assumes failure is always the vendor's fault, when a lot of it comes down to how the customer configured and ran the agent.
• Analysts are split on whether Salesforce's Agentforce revenue is durable, even while looking at the same numbers: a useful reason for healthy scepticism towards any vendor's AI roadmap claims.
• Build-your-own and outcome-based pricing are really the same bet: ongoing platform upkeep is easy to underestimate, whoever's doing it.
• The Caribbean doesn't yet have a homegrown CRM platform on the open market, but demand for custom local builds is growing, which raises the same maintenance question closer to home.

If you're weighing an AI-agent pitch from a CRM vendor right now, happy to talk through what's actually worth testing before you commit to anything.

FAQ

• Is CRM AI pricing moving away from per-seat licensing? Not broadly yet, but Sierra's outcome-based model, paying only when an AI agent resolves an issue rather than per seat, is the clearest current challenge to how HubSpot, Salesforce and other CRM vendors price their AI features today.

• What percentage of enterprise teams have replaced a SaaS tool with something they built themselves? A 2026 Retool report found 35% of enterprise teams have already replaced at least one SaaS product with an in-house build, and 78% plan to build more this year.

• Is Salesforce's Agentforce revenue growing? Yes. Salesforce reports Agentforce annual recurring revenue near $1.2 billion, with combined AI-and-data revenue approaching $3.4 billion, both growing at double-digit rates year over year. Analysts remain split on whether that growth is durable.

• What is outcome-based AI pricing? A pricing model where a vendor charges only when its AI agent delivers a result, whether that's resolving a support ticket, saving a cancellation or closing an upsell, rather than a flat fee per user seat.

• Is there a CRM platform built in the Caribbean? Not yet, as far as we're aware, on the open market. There's growing demand for custom-built CRMs from local providers instead, though that still requires dedicated local capacity to configure, support and keep developing it over time.

Sources

Build-vs-buy / enterprise SaaS replacement:

• Thinking About Replacing Salesforce With Your Own Vibe-Coded CRM? Read This First — Aquiva Labs
• 5 Simple Reasons We Won't All Vibe Code Our Own HubSpot or Salesforce — SaaStr

Sierra's outcome-based pricing:

• AI agents could change how software companies get paid, Sierra co-founder says — CNBC
• AI agents could change how software companies get paid, Sierra co-founder says — AI Commission

Salesforce / Wall Street analyst split:

• Salesforce Inc Stock (CRM) Moved Up by 5.22% on Jul 13: Drivers Behind the Movement — TradingKey
• CRM Stock Juggles Defense Wins, AI Risks, And Analyst Split — StocksToTrade
• CRM Extends Defense Cloud Reach As Wall Street Stays Split — Timothy Sykes