How AI is forcing SaaS leaders to rethink who they serve, how they serve, and what success potential really means.

Most teams think their ICP problem is a sales problem.

The real issue is this: the assumptions you built your ICP on no longer hold. AI is changing cost to serve, segmentation, and value delivery faster than most SaaS companies can update their spreadsheets.

If you don’t revisit ICP now, AI won’t save you. It will simply make your old mistakes scale faster.

Let’s break this down.

The Old ICP Was Built for a Different Reality

For years, ICP was a capacity management tool.

We picked an ICP because:

Our onboarding capacity was limited Our CSM bandwidth was limited Our data on success patterns was limited

So we created a slide. We added some nice stats. We tried to get sales to follow it.

And like we said in the podcast episode: “Sales keeps bringing in anything but ICP.”

Behind that frustration is the fear every CS leader knows: “If we bring in the wrong customers, the whole system breaks.”

That used to be true.

With AI in the stack, it’s less true. But it’s also more dangerous to ignore.

AI Breaks the Old Tradeoffs

AI is forcing us to rethink ICP at a fundamental level. Three shifts matter most.

1. Cost to Serve Drops

A CSM can now cover more accounts with AI handling monitoring, nudges, summaries, next steps, and risk detection. This pushes you to think: Maybe we can go wider.

2. Fit Signals Get Sharper

You no longer rely on static company data to define ICP. You can analyze patterns across onboarding velocity, product usage, expansion behavior, and renewal probability. This pulls you toward: We should go narrower.

3. You Can Personalize Across Segments

You can run divergent motions with the same team: High-touch for strategic customers. Tech-touch for long tail. Self-serve for micro segments.

This means your ICP is no longer a blunt instrument. It can be precise, dynamic, and tiered.

So the real question is not:

“Should we widen or narrow our ICP?”

It’s:

“Where do we go narrow, and where do we go wide - powered by AI and systems, not heroics?”

From ICP to “AI Informed Success Potential”

In the podcast, Lincoln Murphy and I talked about something far more useful than ICP alone:

Success potential.

A simple, binary checklist:

Do they have the right internal owner? Do they have the right tech stack? Do they have the minimum data? Do they have the urgency? Do we have the resources to support them?

If they don’t meet these minimum requirements, they don’t have success potential.

In an AI world, this becomes even more powerful.

Instead of a hand-crafted ICP, you move toward:

Dynamic, AI-informed success potential scoring.

Examples:

Predictive fit scoring based on historical NRR patterns Early onboarding risk signals instead of month 6 surprises Recommended plays based on what great-fit customers actually did A data-driven way to validate your ICP, not guess it

This is where AI drives real revenue impact.

Not in generic chatbots. In fit intelligence and success potential prediction.

So… Narrower or Wider?

Here’s the answer I share with CEOs in boardrooms and workshops:

Narrow your ICP in strategy. Widen your coverage in execution.

1. Narrow in who you actively pursue

Use AI to define the 10–20 percent of the market with the highest historical success potential and NRR upside. This is where sales and marketing spend real energy.

2. Go wider in who you can responsibly serve

AI makes it possible to profitably serve adjacent segments with lighter motions. Not ICP. But not no-go either.

3. Be ruthless with bad fit

Every SaaS CEO says they do this. Almost none actually do.

In an Impact Academy session, a CS leader realized 30 percent of their base had no chance of hitting value. The ICP problem wasn’t top-of-funnel. It was the lack of courage to say no.

AI will surface the truth whether you like it or not.

The Piece Most Teams Miss: Your Own Readiness

Success potential is not just about the customer.

It’s about you.

You have to ask:

Do we have the onboarding bandwidth for this segment? Do we have the content and plays that actually work? Do we have the data quality to model fit accurately?

Early stage companies need a tighter ICP because every account requires heavy lifting. AI helps, but it doesn’t erase the grind.

Later stage companies can go multi-tier and multi-motion with confidence.

AI doesn’t change readiness. It just reveals where you’ve been pretending.

How to Start on Monday

If you want to modernize your ICP with AI, keep it simple.

Step 1: Define success potential 5–10 binary criteria. No AI required. This becomes your ground truth.

Step 2: Tag current customers Great fit. OK fit. Bad fit. Look at NRR, expansion, support load, onboarding time.

Step 3: Layer AI on top of reality Use your data to validate patterns, not to invent theory.

Step 4: Decide explicitly Narrow ICP for active pursuit Wider coverage with AI-supported motions A red zone you actively say no to

This is the new ICP discipline. This is the path to hyper efficient and hyper customer centric growth.