A function can become faster while the business stays stuck. The next customer outcome may depend on decisions that cross several budgets, targets and professional boundaries.
TIP’s argument for cross-functional growth starts there. Use some of the capacity released inside functions to work on a customer problem that those functions cannot solve independently. The purpose is better decisions and a more useful offer. The team is a means to that end.
1. Start with the problem between functions
Consider a customer who wants to adopt your product but cannot get it working quickly. Sales promises an outcome. Product builds features. Operations manages implementation. Customer teams handle the frustration. Each function can improve its own productivity without removing the delay the customer experiences.
A useful growth question is: what would have to change across this journey for a customer to choose us, succeed sooner or stay for a better reason? That question creates a shared unit of work. “Explore AI opportunities” usually does not.
The same logic applies when customers become more capable. If they can bring routine work in-house, Revenue may notice the exposure first. A better offer can require Product to change the service, Operations to change delivery, Finance to change the economics and Legal to clarify the promise.
2. Treat the evidence as a challenge to team design
The P&G experiment published in Organization Science in June 2026 found that AI-assisted individuals could match unaided teams on early product-innovation proposal quality. That challenges the assumption that adding colleagues is automatically the best way to combine expertise.
Teams with AI were more likely than unaided individuals to produce a top-decile solution, but their advantage over unaided teams on that measure was not statistically significant. This was an innovation task in one company, not a test of realised revenue growth. Read the study, scope and limitations →
Our interpretation is practical: choose the smallest team that adds something an individual cannot easily supply. That might be customer knowledge, technical understanding, delivery authority or the ability to commit resources. Ask what each person changes in the decision. Do not use the study to justify a standing committee.
The published study also distinguishes generating ideas from selecting them: AI improved idea generation, while human judgement retained value in selection. Our interpretation is that a team needs a reason to exist beyond producing more options.
3. Pool capacity, then give it somewhere to go
The 10·10·10·10 model illustrates four adjustable buckets. Ten percentage points for growth is one possible choice, not a prescribed allocation. You may hold headcount flat and give all released capacity to risk and growth. The functions create capacity; the leadership team collectively decides its destination against strategy. Explore the two stages and adjust the model →
Across functions, that time can support a useful experiment. It will not do so if everyone keeps the old workload and treats growth as a side task. A separate 66-firm experiment found changes in email time without detecting a broader reallocation of tasks from individual tool provision. Saving time and changing work are different management problems. Read the evidence →
4. Change the operating conditions
A team drawn from several functions is not automatically working differently. The conditions below matter more than the label.
| Make explicit | What it changes |
|---|---|
| One customer outcome | People optimise the same result rather than defend separate outputs. |
| A named sponsor | Someone can resolve a trade-off that sits beyond the team’s authority. |
| Decision rights | The team knows what it can change, spend, promise and stop. |
| Protected time | The work has capacity because something else has been removed or deferred. |
| Shared evidence | Customer observations, economics, feasibility and risk are considered together. |
| A review date | The experiment ends in a decision, rather than becoming permanent activity. |
Incentives deserve particular attention. A seller rewarded only for immediate bookings may resist a smaller test. An operations manager measured only on utilisation may resist reserving capacity. A product team committed to feature throughput may treat service changes as a distraction. The sponsor must make these conflicts visible and decide which outcome takes priority.
5. Run a six-week learning cycle
The timetable below is a proposed starting point. Shorten or lengthen it to fit the uncertainty and the customer relationship.
- Week 1: frame. Pick one customer problem, gather current evidence and name the assumption most likely to invalidate the opportunity. Establish a baseline and decision rights.
- Weeks 2–3: make a small test. Prepare the minimum offer, prototype or service change needed to test that assumption. Agree the cost ceiling, customer expectations and stop conditions.
- Weeks 4–5: observe. Work with an appropriate, limited customer group. Record what people actually do, delivery effort, quality and unintended effects. Distinguish a promising conversation from a purchase or successful outcome.
- Week 6: decide. Stop, change the hypothesis or fund another bounded step. Keep the evidence even when the idea fails.
Measure the relevant customer outcome alongside economics and delivery. Examples include successful onboarding, verified willingness to pay, repeat use, reduced avoidable contact or retained contribution. A prototype count or the number of AI prompts is not evidence of growth.
6. Know when a team is unnecessary
Use an individual with appropriate review when the work is bounded, the needed knowledge is available and decisions sit with one owner. Use a temporary pair when two perspectives are sufficient. Convene a wider team when the opportunity depends on several interdependent decisions and those people can actually act on them.
AI can help assemble information, challenge a proposal and explore another professional perspective. It does not automatically supply lived customer knowledge, authority to commit the business or accountability for the consequences.
One thing to do: write the experiment charter
In a 45-minute session, choose one customer problem that has repeatedly crossed functional boundaries. Complete the following seven lines together. If you cannot name a specific outcome or a decision the team can make, reduce the scope before adding people.
- Customer and problem: who struggles, with what, and what evidence do we have?
- Outcome: what observable improvement would matter to that customer?
- Assumption: what must be true for this to become a worthwhile offer?
- Smallest team: whose knowledge or decision rights are indispensable?
- Capacity: what will each person stop or defer, and for how long?
- Boundaries: what can we spend, change and promise; what would stop the test?
- Decision: when will we review the evidence, and who decides the next step?
The output is one page, one sponsor and one learning cycle. That gives released capacity a destination—and gives the organisation a way to discover whether its next opportunity requires it to think and work differently.
video podcast · 60 min 19 sec · Start at 37:58
Ethan Mollick: How to build an AI-first organization
Sana · Strange Loop
A researcher explains an experiment that challenges assumptions about individuals, expertise and teams. Start at 37:58 and ask what a team must add beyond a capable individual with AI.
Listen or watch for: If AI helps individuals cross expertise boundaries, what additional value should a human team create?
About this resource
One study’s author discussing a specific experiment, not proof that every task needs a cross-functional team. The page warns its AI transcript may contain errors.