The practical question is not whether your company should become full-stack.

The question is where ownership would actually improve the system.

This audit is designed for a leadership team deciding what to own, buy, partner for, outsource, automate, or leave alone in the AI era. Use it for a product line, business unit, or company strategy review.

Score each section from 0 to 2. 0 means unclear or weak. 1 means partially designed. 2 means explicit, owned, and supported by evidence. The score matters less than the gaps it exposes.

Use the audit in a working session, not as a survey. The value comes from disagreement: product may think the workflow is owned, services may know the exceptions are unmanaged, finance may see margin risk, and sales may know trust is the real blocker. Put those differences on the table.

1. Customer outcome clarity

Start with the result.

Ask:

  • What outcome does the customer actually want?
  • Is the current product a tool, workflow, managed service, or outcome promise?
  • What must happen before the customer receives value?
  • Where does value fail today?
  • Which parts of the outcome are under our control?
  • Which parts depend on the customer or third parties?

If the outcome is vague, do not make integration decisions yet. You will not know which layers matter.

2. Workflow ownership

Map the workflow that produces the outcome.

Identify:

  • steps;
  • owners;
  • tools;
  • handoffs;
  • approvals;
  • exceptions;
  • quality checks;
  • data created;
  • feedback loops;
  • failure points.

Then ask: where would owning more of this workflow improve speed, quality, trust, learning, or economics?

If the answer is nowhere, vertical integration is probably unnecessary. If the answer is obvious in one or two places, focus there.

3. Data-loop strength

Do not ask only whether you have data. Ask whether you have a loop.

For each important workflow, inspect:

  • context captured before action;
  • AI recommendations or system outputs;
  • human edits, approvals, rejects, and escalations;
  • outcome labels;
  • freshness;
  • permissions;
  • owner;
  • improvement mechanism.

A strong data loop turns work into better work. A weak one produces logs nobody uses.

4. Service layer design

If services are involved, classify them.

Which service work is:

  • strategic human judgment;
  • transitional manual work;
  • implementation glue;
  • exception handling;
  • non-strategic customization?

For each category, define whether the work should remain human, become productized, become automated, be priced separately, be handled by partners, or be refused.

Services are healthy when they teach the system. They are dangerous when they hide product gaps forever.

5. Distribution control

Assess how the company reaches and keeps the customer.

Ask:

  • Who owns the customer relationship?
  • Which channels create trust?
  • Where does adoption friction show up?
  • Who captures product and workflow feedback?
  • What distribution is rented from partners or platforms?
  • What would happen if a major channel changed rules or disappeared?

Distribution is part of the stack when it creates access, trust, data, or learning competitors cannot easily copy.

6. Trust and compliance

For sensitive workflows, audit trust explicitly.

Ask:

  • What data or decisions create risk?
  • What must be auditable?
  • What permissions are required?
  • What compliance obligations apply?
  • What buyer concerns block adoption?
  • What evidence can sales, legal, security, and product produce today?
  • What failures would damage the brand?

If trust is required to access the workflow, trust is not a side function. It is product infrastructure.

7. Economics

Build the economic picture before committing.

For each candidate layer to own, estimate:

  • revenue upside;
  • willingness-to-pay impact;
  • gross margin today;
  • gross margin after productization or automation;
  • labor needs;
  • capital needs;
  • utilization risk;
  • cost-to-serve by segment;
  • retention or expansion impact;
  • dependency-risk reduction;
  • management complexity.

If the model works only under heroic assumptions, treat that as a strategy problem, not a spreadsheet problem.

8. Capability fit

Ask whether the company can operate the layer well.

Some layers require capabilities the company does not have: field operations, regulatory expertise, service management, hardware, security, financing, data operations, partner management, or enterprise implementation.

That does not mean the company should avoid them. It means the capability build must be explicit.

A layer can be strategically correct and operationally premature.

9. Non-integration decisions

Name what you will not own.

For each layer in the broader stack, decide:

  • own;
  • buy;
  • partner;
  • outsource;
  • automate later;
  • leave alone.

Write the rationale. This prevents strategy from becoming a one-way ratchet toward complexity.

The strongest full-stack companies have sharp noes.

10. The 90-day action plan

End with decisions, not vibes.

Choose:

  • one workflow to instrument better;
  • one data loop to strengthen;
  • one service activity to productize or stop;
  • one distribution dependency to reduce or manage;
  • one trust layer to improve;
  • one economic assumption to validate;
  • one layer explicitly not to own.

Assign owners and review dates.

Also name one decision you are deliberately deferring. A good audit should create action, but it should also prevent premature integration where the evidence is not ready.

Reading the score

After scoring, look for patterns:

  • Low outcome clarity means stop and sharpen positioning before integrating.
  • Low workflow or data-loop strength means instrument before automating more.
  • Low service design means productization is probably blocked by unmanaged delivery.
  • Low distribution or trust scores mean access, not capability, may be the bottleneck.
  • Low economics or capability fit means the strategy may be right but premature.

The next move should address the binding constraint, not the most exciting layer to own.

The executive takeaway

AI makes vertical integration more attractive because more value lives in loops that cross product, data, workflow, services, trust, and distribution.

But integration is not the goal. Better outcomes, faster learning, stronger trust, improved economics, and strategic control are the goals.

Own the layers that compound. Refuse the layers that merely make you feel in control.

That is the full-stack company discipline.