6 Signs an AI SDR or GTM Vendor Won't Be Around When You Need It

Why I'm writing this list now

I'm Hlib Storchak. I build outbound systems for B2B founders and sales teams, and I've booked 2000+ meetings for B2B clients doing it, which means I've also had to rip out and rebuild pieces of a client's stack when the tool underneath it quietly disappeared. On April 1, 2026, Persana AI, a B2B sales automation platform with an autonomous SDR agent called Nia, announced it was "joining forces" with Rox. The standalone Persana platform was sunset by May 2, per Persana's own migration guidance and independent trackers, with a roughly 30-day window to export data before it was gone for good. If you had built a research or enrichment workflow around Persana, you had about a month's notice, packed into a company blog post titled like good news.

This isn't a one-off. A running changelog of 2026 B2B data and sales tool acquisitions and shutdowns kept by Cleanlist.ai lists Pocus folding into Apollo.io, Warmly being acquired by HubSpot, Common Room closing into Zoom, Qualified joining Salesforce, and Seam AI folding into Clarify, all within about five months of each other. None of that makes AI GTM tooling a bad category to build on. It does mean vendor continuity deserves the same due diligence you'd give a vendor's accuracy claims, and right now most buyers only check the second thing.

How I'm ranking these six signs

I ranked these by how much real warning they actually give you, not by how alarming they sound. A sign that shows up 6 to 12 months before a shutdown is more useful to you than one that only makes sense in hindsight, which is why the list below runs roughly from "check this quarterly" to "check this once, at signup, and move on." I'm using real, named examples wherever I found one I could verify directly. Where I couldn't verify a specific number, I say so instead of repeating it, which is the same rule I hold vendors to elsewhere on this blog.

1. The roadmap goes quiet, then a "next chapter" post shows up

Persana's own announcement didn't use the word "shutdown." It used "joining forces" and "next chapter," and it didn't even commit to a specific sunset date in the post itself, only "over the coming weeks." The concrete dates, an April 1 to May 1 export window and a May 2 deletion, came from Persana's separate migration guide and were confirmed independently by third-party trackers, not from the announcement post a customer would actually see first. Verdict: the language softens the timeline every time. Read "next chapter" as "shutdown," and go find the actual date in the migration doc, not the blog post.

2. The shutdown notice lives on someone else's site, not the vendor's own

Clado, a people-search tool, wound down and handed its requests to a successor called Clodo. Per Cleanlist.ai's tracked changelog, the wind-down notice appears on Clodo's site, not on Clado's own properties. I could not find a first-party wind-down notice on Clado's own site either, which is itself the point: if the only place you can confirm a tool is going away is a competitor's or successor's blog, you've already lost whatever notice period you thought you had. Verdict: if you can't find the news on the vendor's own domain, assume you're finding out later than you should have.

3. The vendor's urgency stat doesn't survive a source check

A specific number is circulating right now across AI-agent blogs: some version of "88% of AI agent pilots never reach production," sometimes credited to Forrester and Anaconda, sometimes to Gartner. Checking further, the number moves depending on which site you land on, 88%, 89%, 78%, 86%, all attributed to the same rough claim with no two sites citing an identical, traceable source. Gartner's own newsroom does carry a real, dated prediction in this territory, a June 25, 2025 press release stating over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating cost, unclear business value, and inadequate risk controls. That is a real, checkable number, and it is not 88%, 89%, or any of the other figures I found attributed to Gartner elsewhere. I've already built a full framework around that real 40% figure in a separate piece, so I won't re-litigate it here. The point for this list is narrower: when a vendor's sales deck or a blog post cites a scarier, rounder number to create urgency around signing now, that's worth an extra five minutes before it moves your decision. Verdict: a vendor or aggregator citing an unsourced pilot-failure stat to rush your decision is a sign to slow down, not speed up.

4. Pricing changes more than once a year with no explanation

HubSpot moved its Breeze Prospecting Agent to $1.00 per qualified lead in April 2026, part of a broader shift toward outcome-based AI pricing. Salesforce's Agentforce SDR Agent bills through metered credits tied to action volume rather than a flat rate, and reviewers covering it note the pricing mechanic itself has already been adjusted more than once since launch. I go deeper on exactly how those two pricing models compare in a dedicated comparison. Repricing on its own isn't damning, every vendor tunes pricing as it learns its own unit economics. What's worth watching is the pattern: a vendor that changes its pricing mechanic, not just the number, more than once within a year is usually still searching for a model that actually works for them, and that search sometimes ends in exactly the kind of "next chapter" post that opened this list. Verdict: one repricing is normal. A different pricing mechanic every few months is the vendor telling you, indirectly, that they haven't found solid ground yet either.

5. One team, one round, and a roadmap that hasn't shipped in a year

Gojiberry AI, a YC-backed AI GTM agent founded in 2025 by three repeat founders, markets an intent-signal-driven outreach agent and claims teams see "3 to 5x more replies," a self-reported figure with no disclosed methodology I could find, which I'm noting here as a claim, not a fact. I'm not picking on Gojiberry specifically, and I have no reason to think it's in trouble. I'm using it, alongside where Persana started, to make a general point: a small team on a single funding round, still in year one or two, carries real continuity risk that has nothing to do with product quality. That's not a reason to avoid young vendors. Plenty of the best tools I recommend started exactly this way. It's a reason to price that risk into how deeply you integrate with one before it has a longer track record. Verdict: young and small isn't a red flag by itself. Betting your whole prospecting-to-booking loop on a year-old, single-round vendor without a fallback plan is.

6. No independent reference customer after 12+ months in market

I cover how to actually run a reference check in more depth in a separate piece on vetting an AI SDR vendor's claims, so I'll keep this one short: if a vendor has been generally available for over a year and still can't produce a customer willing to talk to you directly, that's not a sample-size problem, it's a signal about how many customers actually stuck around long enough to become a reference. Case studies with a logo and a quote are not the same thing as a name and a phone number you can call yourself. Verdict: no reachable reference after a year in market tells you more about retention than any number on the pricing page does.

All six signs, ranked side by side

What to put in the contract before you sign

A few clauses do more for you here than any amount of vendor-watching, because they turn "hope the vendor gives you notice" into an obligation on paper:

• A minimum notice period before termination or sunset, 60 to 90 days rather than the "coming weeks" language Persana's customers got.
• A guaranteed data export window and format, a full CSV or API export of your own data, not just what the vendor's UI happens to expose that week.
• No auto-renewal without a break clause, so a shutdown decision on their end doesn't collide with a locked annual term on yours.
• A decoupled infrastructure layer. This is one reason I keep sending infrastructure, domains and mailboxes, on its own layer, separate from whatever AI research or enrichment tool sits on top of it. I run Mailforge and Infraforge for that specifically, for my own stack and for clients. If the enrichment or AI layer disappears, your domains and warmed mailboxes are untouched, and you're swapping one component, not rebuilding the whole engine.

What a forced 30-day migration actually costs

Nobody publishes a real number for this, so here's a model built on stated assumptions you should swap for your own. Assume a mid-market team running a $400 to $800/month AI GTM tool that's wired into your CRM and enrichment waterfall, not just sitting off to the side. A realistic rebuild: an ops person spending 15 to 25 hours re-mapping integrations and re-testing data flows at roughly $45/hour, plus a founder or sales leader spending 4 to 6 hours evaluating and selecting a replacement, plus a real pipeline gap while sequences pause, call it one to two weeks of reduced outbound volume.

That's before pricing in whatever a one-to-two-week pipeline gap costs in meetings not booked, which depends entirely on your own funnel numbers. Swap in your own hourly rates and your own integration depth. The shape of the model, not my specific figures, is the part worth keeping: the cost of a forced migration scales with how deeply the tool is wired into the rest of your stack, which is exactly why the decoupling advice above pays for itself the first time you need it.

The mistake I see most often

The mistake I see most often when I take over a client's stack isn't that they picked a bad vendor. It's that they never wrote down which parts of their outbound engine depend on which vendor, so when one disappears, the first hour is spent figuring out what actually broke before anyone can start fixing it. A simple dependency map, which tool owns sending, which owns enrichment, which owns signals, which owns the CRM record, takes maybe 30 minutes to build and turns a scramble into a known, bounded swap. I build that map for every client engagement now, before we touch a single sequence, specifically because of days like the one a lot of Persana's customers had in April.

Where I land

None of this is an argument for sticking only with the biggest, oldest vendors. Some of the best tools I've recommended on this blog are young companies taking real product risks that the incumbents won't. It's an argument for treating vendor continuity as its own line item in due diligence, separate from whether the product works and separate from whether its case studies check out. A tool can be genuinely good and still not be around in 18 months. Plan for that possibility explicitly, with a contract clause and a decoupled architecture, and a shutdown announcement becomes an annoying Tuesday instead of a scramble.

FAQ