Is your SaaS AI-resilient? Seven questions to ask before your next roadmap meeting

Almost every founder I talk to has had the same uncomfortable thought in the last year: could a model just do what we do? Some shrug it off. Some rebuild their roadmap around a chat box. Neither is a strategy.

AI resilience isn't about how much AI is in your product. It's about whether your product still matters when anyone can get a decent version of your output from a prompt. Here's how I think about it, as the founder of a company that sits right in the middle of this (an analytics product in a world where everyone can ask a model to "analyze my revenue").

Two questions that sort almost everything

When you strip it down, two things decide how exposed a SaaS product is:

1. How much of your value is an output a model can generate? Text, images, summaries, simple analysis and first drafts are all things a general model now does well and cheaply.
2. How deeply are you embedded in the customer's data and workflow? Are you where the work happens, where the records live, and where the numbers have to be right?

Products that mostly generate output and sit outside the workflow are the most exposed: that's where a prompt can replace a subscription. Products that are deeply embedded and don't depend on generated output are the safest. The interesting part is the two middle quadrants, because that's where most SaaS companies are, and where the next two years of roadmap decisions matter.

The seven questions

1. Could a customer get 80% of your core value from a prompt? Be honest. If the answer is yes, your product isn't doomed, but your moat has to be something other than the output: the data, the workflow, the trust, or the distribution.

2. Do you have data the model doesn't? General models know the internet. They don't know your customer's billing history, their support tickets, or three years of their usage. Data that accumulates inside your product, and gets more useful the longer a customer stays, is the strongest defense there is. It's also what makes AI features inside your product better than a general chatbot.

3. Are you where the work happens? A system of record, a workflow people run every day, a tool connected to five other tools: these are hard to replace, because replacing them means changing how a team works, not just where they get an answer.

4. Is being right your job? AI makes plausible answers cheap. That makes verified answers more valuable. In our world, a model can happily invent an MRR number. The value of GrowPanel is that every number is rebuilt from the customer's actual billing history and reproducible. If your product is the place where the numbers, the compliance, or the audit trail have to be correct, AI raises your value rather than lowering it.

5. Can agents use you? More and more work will be started by an AI assistant rather than a person clicking through your UI. If your product can't be called by an agent, it gets routed around. An API, good docs, and a way for assistants to connect (like an MCP server) turn AI from a threat into a distribution channel. We added a hosted MCP server for exactly this reason: people now ask Claude or ChatGPT about their MRR, and we want the answer to come from GrowPanel.

6. Does your pricing survive fewer seats? If AI means your customers need fewer people to do the same work, per-seat pricing shrinks even when your value grows. It's worth modelling what happens to your revenue if your average customer has 30% fewer seats in two years. The academy has a primer on per-seat pricing and its alternatives.

7. Do you own your distribution? If most of your customers find you through search, and search is increasingly answered by AI assistants, your acquisition channel is changing under your feet. Brand, community, partnerships and being the answer that assistants recommend all matter more than they did.

A quick scorecard

Score yourself from 0 to 2 on each question and add it up. It's crude, but it makes the discussion with your team concrete.

A low score doesn't mean you should panic. It means the next roadmap should move you right and up on the matrix: capture more of the customer's data, get into the daily workflow, make being correct your selling point, and make yourself easy for agents to use.

Resilience shows up in your metrics first

You won't see AI disruption in a headline. You'll see it in your retention numbers, months before it shows up in total MRR. Watch these:

• Gross revenue retention. If customers start leaving for "we'll just use ChatGPT for that", GRR drops first.
• Contraction. Customers downgrading or removing seats is often the step before churn, and it's the first place fewer seats will show up.
• Cohorts. Compare customers who signed up this year with those from two years ago. If newer cohorts retain worse, something in the market has changed. Cohort analysis makes that visible.
• New customer mix. If your new customers skew towards a segment where the AI alternative is weaker, that's a sign of where your real moat is.

GrowPanel breaks your MRR into new, expansion, contraction and churn every month and shows retention by cohort, so these shifts are visible early. If you want to see what that looks like, take a look at the live demo, or read how SaaS CEOs use GrowPanel to spot risks before they show up in the total.

The companies that come out of this stronger won't be the ones with the most AI features. They'll be the ones customers can't imagine working without, with or without a model in the loop.