Why "Soft Pivots" Always Become Two Companies in One Office

I want to describe a pattern I've seen at three companies in three industries in the last eighteen months. Different products. Different scales. Same outcome.

The CPO and CEO decide they want to "evolve toward an AI-native product." Not cannibalize. Not pivot. Evolve. Gradual. Customer-friendly. Disciplined.

They keep selling the legacy SaaS at full price. They start building agent-native capabilities alongside it. They hire a few "AI specialists." They run a couple of pricing experiments on the agent side. They have a slide that says "we are bringing AI to our category" with two product lines on it that look complementary.

Eighteen months later, both product lines are slow. The legacy product's NRR is softer than expected. The agent product's revenue is much smaller than the plan. Engineering is exhausted. The CRO and the CPO are tense with each other. The board is asking what the pricing strategy actually is.

This is not bad luck. This is the predictable outcome of the soft pivot. And the only way to escape it is to admit you're running it.

The short version

The soft pivot becomes two companies in one office because it has two cadences, two pricing models, two talent profiles, two customer segments, and one shared P&L. The shared P&L hides the operational tension. Every shared resource (engineering attention, sales motion, CS capacity, customer trust) becomes a turf war. The org chart pretends it's one company; the work is two. Within twelve months, both halves are slow, and the company concludes "we're not good at AI." It wasn't an AI problem. It was a sequencing problem.

The fix is to publish a sunset on the legacy. The Cannibalization Playbook walks through the strategy. The Dual Transformation Operating Model walks through the rhythms. This essay is the diagnostic. If three of the patterns below are happening in your company, you are inside a soft pivot, and the cost of staying inside it compounds every quarter.

The five patterns

Pattern 1: Engineering attention is "shared" between the two products

When you ask the head of engineering how engineers are allocated, the answer is something like "the team works on both. We try to balance." The reality is engineers are constantly context-switching. They start the morning on a legacy bug and end the day on an agent eval review. They feel productive in neither. Velocity drops on both products.

If your engineers can't tell you which product they're "on" in one sentence, you are running two companies on shared engineering, and shared engineering is no engineering at all.

Pattern 2: The pricing page has both models with no clear forward path

Your pricing page shows per-seat plans and outcome plans, both. Sales reps explain "we have flexibility" to prospects. Customers ask which one they should pick and the answer depends on the rep. Some accounts get hybrid pricing that nobody at the company can explain.

If your own salespeople can't tell prospects what the right choice is, you don't have two pricing models. You have one pricing model that's drifting plus a side experiment, and the experiment is being overshadowed by the legacy because the legacy is what gets paid on.

Pattern 3: NRR conversations and outcome conversations happen in different rooms

The NRR review meets monthly. The outcome cohort review meets weekly. The two meetings have different attendees and different vocabularies. The NRR review treats the agent product as a small line item. The outcome review treats the legacy as a constraint. Nobody is reconciling them.

If the two halves of your company aren't in the same room reviewing performance together, the gap will grow. The legacy team will start to think the agent team is impractical. The agent team will start to think the legacy team is dragging them down. Both will be partially right. The company will be wholly wrong.

Pattern 4: Customer escalations cross the seam

A legacy customer files an escalation about an agent feature you bolted on. The legacy CS team doesn't know how to handle it. They forward to the agent team. The agent team doesn't know the customer's history. They forward back. The customer waits. The customer churns.

Customer escalations crossing the product seam are the canary. Every time it happens, you are paying a tax for running two products as one. The tax is invisible on the P&L and lethal on retention.

Pattern 5: The board update has two slides that don't add up

The legacy slide shows revenue, NRR, CAC, gross margin. The agent slide shows outcome volume, dispute rate, gross margin per outcome, customer references. Neither slide has the migration percentage. Neither slide says "this is the integrated transition story."

If your board doesn't know which number to ask about, neither do you. The slides are reflecting an org that hasn't decided what it's doing.

Why the pattern is invisible from inside

The soft pivot is hard to see from inside the company because it looks like normal product growth on both sides. Each individual decision is reasonable. The legacy team should keep selling. The agent team should keep building. The pricing experiments should run. The CS team should help everyone. Of course.

What's invisible is the accumulating drag of having those reasonable decisions stacked on top of each other without a sequencing call.

Here's the test that works. Write down, on one page, the answer to these five questions:

1. What's the sunset date for the legacy product?
2. Who's the engineering lead exclusively on the agent product, with no rotation?
3. What's the comp asymmetry between selling legacy and selling agent?
4. What number does the CEO track every week that integrates both products?
5. What's the agreed gross margin trough and the planned recovery?

If you can't answer all five in one sentence each, you're in a soft pivot. The seam is real even if you can't see it.

The way out

The way out is unglamorous and well-understood. Set the sunset. Reorganize. Rewrite comp. Pre-sell the trough. The strategic structure is in The Cannibalization Playbook. The operational structure is in The Dual Transformation Operating Model.

What I want to leave you with here is the thing that holds most CPOs back from doing it.

The reason CPOs prefer the soft pivot is that it doesn't force any individual hard call. No customers told their product is sunsetting. No engineers told they're on the maintenance team. No board update titled "We are killing our own product." The soft pivot is the strategy of avoiding any single moment of pain.

The cost of avoiding the moment is paying it in installments over three years. And the installments add up to more than the lump sum.

I've watched a CEO friend describe his soft pivot as "preserving optionality." Eighteen months later he laid off 30% of the company because the legacy revenue had collapsed faster than the agent revenue could grow, and he was out of runway. The optionality he preserved was the optionality of the inevitable.

If you have to lay off 30% of the company at the end of the soft pivot, you should have laid off 5% at the start of the cannibalization. Same math. Different timing. Different reputation.

What to try this week

Take the five-question test and answer it honestly on one page. Show it to one trusted peer. Ask them to push back on each answer.

If you can't answer all five, you have your week-of-work cut out for you. The first conversation is with the CFO, not the team.

The discipline of the cannibalization is what separates the executives who navigate this inflection from the executives who get displaced by it. The soft pivot is the comfortable choice. It is not the safe one.

---

This essay sits inside the /cpo executive lane. The Cannibalization Decision Tree (a single-page worksheet for running the five-question test) is at /toolkit/cannibalization-decision-tree.

Further reading

• The Cannibalization Playbook
• The Dual Transformation Operating Model
• Adobe's Creative Cloud transition (Stratechery)
• Forrester: Predictions 2026