AI is breaking the SaaS assumptions behind activation, retention, and expansion. Here’s what growth teams should measure instead.
By Zain Abiddin, Founder of GrowthPad
For the last 15 years, SaaS companies have been optimizing roughly the same growth machine:
Acquire users. Get them to sign up. Move them through onboarding. Drive feature adoption. Convert them to paid. Retain them. Expand the account.
We built an entire growth discipline around those steps—and the metrics that represent them: activation rates, daily and monthly active users, feature adoption, seat expansion, and net revenue retention.
But AI is starting to break a fundamental assumption underneath that model.
Customers increasingly do not want to use more software. They want software to do more of the work.
I do not necessarily want another analytics dashboard. I want to know why revenue dropped.
I do not want another CRM workflow builder. I want qualified prospects followed up with.
I do not want 20 new AI features. I want the job done.
If that is where software is heading, we need to rethink what growth actually means.
The question is no longer just: How do we get customers to use more software?
It is becoming: How do we help customers accomplish more while using the software less?
The hidden assumption in the SaaS growth model
The traditional SaaS model assumes that the customer operates the software.
The customer logs in, creates things, clicks buttons, configures workflows, invites teammates, and performs actions. Those actions eventually produce an outcome.
So growth teams learned to measure the actions:
Did the user create three projects?
Did they invite five teammates?
How many features did they use?
How many sessions did they have?
What is the ratio of daily to monthly active users?
For traditional SaaS products, these were often useful proxies for customer value.
AI creates a strange possibility: the product can get dramatically better while those metrics get worse.
What if engagement goes down?
Imagine a workflow that previously required 30 minutes inside your product. An AI agent can now complete it in 30 seconds.
Sessions decrease. Clicks decrease. Time in product decreases. Feature usage might decrease too.
Yet the customer receives more value.
That creates a dangerous situation for companies still running their business on traditional engagement dashboards. You could improve the product while making your core metrics look worse.
AI can separate engagement from value.
This does not make engagement useless. It makes engagement a proxy—and potentially a much less reliable one than it used to be.
Growth teams need to get closer to the result the customer actually cares about.
From feature adoption to outcome adoption
Traditional product analytics asks:
Which features did the customer use?
How often did they use them?
How many actions did they perform?
For an AI product, I would ask a different set of questions:
Did the customer get a useful result?
Did the product complete the task successfully?
Did the customer accept the output?
Did they actually use the output?
Did they return with another job?
Are they trusting the product with more valuable work?
That is the shift from feature adoption to outcome adoption.
The goal of onboarding is no longer simply to teach someone how the product works. It is to get them to their first meaningful outcome as quickly as possible.
Instead of saying, “Here are the tools you need to solve your problem,” an AI-native product can increasingly say, “Tell me what you are trying to accomplish”—and then do it.
That collapses time-to-value. It also changes what activation should mean.
Trust is the new activation
When someone uses an AI product for the first time, they are not only evaluating whether a feature works. They are deciding whether they can trust the result.
Can I trust this answer? Can I send this output to a customer? Can I give the product access to company data? Can I let it take an action? Can I eventually allow it to work without checking every step?
That is a fundamentally different relationship with software.
“Generated first output” may be an early milestone, but it is not necessarily a meaningful activation event.
Stronger signals might be:
Accepted the first output
Used the first recommendation
Authorized the first action
Reused an AI-generated asset
Delegated a second, higher-value job
Each event tells us that the customer crossed a trust threshold.
Once trust grows, delegation becomes possible.
The Delegation Ladder
AI adoption is not binary. Customers move through increasing levels of trust and responsibility.
I think of this as the Delegation Ladder:
Ask: “Tell me something.” The AI acts like a search or research tool. Risk and trust are low.
Create: “Write, analyze, design, or build this.” The customer begins trusting the output.
Recommend: “Tell me what I should do.” The product begins influencing decisions.
Act: “Send, update, book, deploy, or change this—but ask for approval first.”
Operate autonomously: “Handle it. Tell me when something important happens.”
Every step requires more trust. Every step can also create more customer value.
That gives growth teams a more useful question than “How do we drive adoption of another feature?”
Ask instead: How do we earn enough trust for the customer to delegate more work?
Delegation could become the new expansion
This shift has major implications for monetization.
A large part of SaaS expansion has historically come from seats. The customer hires more employees, buys more licenses, and expands annual recurring revenue.
But what happens when the product’s promise is that the customer needs fewer people to accomplish the same amount of work?
The customer could receive more value while needing fewer seats. If seats are the foundation of your pricing model, your monetization may move in the opposite direction from the value you create.
AI expansion may increasingly come from:
More tasks
More workflows
More agents
More actions
More volume
More valuable outcomes
More autonomy
In other words: more work delegated to the product.
This is why AI pricing and AI growth strategy are becoming difficult to separate. Your monetization model needs to expand with the unit that actually creates customer value.
Context can create a compounding retention loop
Traditional SaaS products build retention through stored data, collaboration, integrations, workflows, and history.
AI introduces another potentially powerful retention mechanism: context.
Over time, an AI system can learn a customer’s company, preferences, previous work, customers, writing style, processes, decisions, standards, and workflows.
The first week may feel generic. Six months later, the product should understand how that customer works.
But storing context is not a moat by itself. The customer has to feel the improvement.
If six months of accumulated context produces essentially the same experience as day one, you have not created much of a retention advantage.
This is more meaningful than reporting that a user logged in 14 times this month. It describes a mechanism through which the product can become more valuable to an individual customer the longer they use it.
That is the type of retention advantage AI companies should be trying to build.
More usage can still produce a worse business
There is another side to this model.
More AI usage does not automatically mean a better business.
Historically, the marginal cost of another query, document, or workflow in a SaaS product could be negligible. AI changes that. Models cost money to run. Agents may make repeated model calls, invoke tools, generate media, run code, search, reason, and execute multiple steps before producing a result.
An AI company can end up in a bizarre position:
Usage is rising. Engagement is rising. Customers love the product. And the unit economics are getting worse.
Growth teams therefore need to care about a metric they could previously leave mainly to finance: cost to serve.
The goal is not to maximize usage. It is to profitably deliver valuable outcomes.
A growth scorecard for AI-native products
If I were building a growth scorecard for an AI-native subscription product today, I would start with seven dimensions:
Intent: Are customers bringing the product meaningful jobs?
Outcome: Is the product solving those jobs successfully?
Trust: Are customers accepting and using the results?
Delegation: Are customers entrusting the product with more responsibility?
Retention: Do they keep returning with more jobs?
Expansion: Is the amount or value of delegated work increasing?
Economics: Can the company profitably deliver those outcomes?
Clicks, sessions, feature count, and time in product may still help diagnose behavior. But they should not be mistaken for value itself.
As AI changes how customers interact with software, the distance between those proxies and actual value may widen.
The new AI growth model
Put everything together and the growth engine starts to look different:
Start with the customer’s intent: What are they actually trying to accomplish?
Get them to a meaningful outcome as quickly as possible.
Use that outcome to build trust.
Turn trust into delegation.
Turn repeated delegation and successful outcomes into retention.
Earn the right to handle more or higher-value work and drive expansion.
Make sure the economics underneath the system work.
Suggested visual: Intent → Outcome → Trust → Delegation → Retention → Expansion, with Economics beneath the full system.
The old growth question was: How do we get people to use more software?
The AI-native growth question is: How do we get software to create more value for people?
Do not copy this funnel blindly
There is one important warning: do not take this framework back to your team and start optimizing all seven stages at once.
That is how growth teams end up doing 100 things and moving nothing.
Your business probably does not have seven equally important problems. It has a constraint.
Maybe retention is excellent, but nobody knows you exist. That is an acquisition problem.
Maybe thousands of people sign up, but few reach a meaningful outcome. That is an activation problem.
Maybe people love the first experience, but never return. That is a retention problem.
Maybe retention is strong, but accounts never expand. That may be a packaging, pricing, or delegation problem.
Maybe AI usage is exploding, but the unit economics are terrible. That is an economics problem.
These are completely different businesses. They should not run the same growth playbook.
More experiments are not the answer
Growth teams love experiments: change the landing page, launch a referral program, redesign onboarding, add annual plans, build an AI feature, start outbound, or run ads.
But an experiment is only useful if it attacks a problem that matters.
If your largest constraint is retention and your team spends the quarter improving website conversion, you might produce a beautiful experiment result. Conversion rises 12%. Everyone celebrates. The company’s trajectory barely changes.
The goal is not to run more experiments.
The goal is to find the highest-leverage constraint and systematically attack it.
That is the discipline behind the GrowthPad Growth OS:
Diagnose the entire growth engine.
Measure where growth is actually breaking down.
Prioritize the highest-leverage constraint—and decide what not to work on.
Experiment with a small number of high-quality bets against that constraint.
Execute those bets through a focused operating plan.
The tactics will change. AI will change. The technology will definitely change.
The discipline of finding the constraint will not.
Two questions for your team
If AI fundamentally changes how customers receive value from your product, has your growth model changed with it?
And what is the single biggest constraint preventing your company from growing faster right now?
If you do not have a confident answer, do not add another disconnected experiment to the backlog.
Map the engine. Find the constraint. Choose the few bets that matter. Build the operating plan.
That is exactly what we do in the GrowthPad Growth Strategy Workshop: a three-hour working session where founders and growth leaders use their own business, numbers, assumptions, and priorities to build a focused 90-day growth plan.
GrowthPad helps founders and growth leaders diagnose their growth engine, identify the constraint, and build a focused plan to fix it. Subscribe to The Growth Roundup for practical frameworks on activation, retention, monetization, expansion, and AI-native growth.
Leave a Reply