Making AI investment decisions across a portfolio

The operating decision

An operating partner needs to decide where the next unit of attention, capital, and implementation capacity should go. A promising use case is only a starting point: the team also needs company context, an accountable owner, a feasible delivery approach, and evidence that the change will be adopted.

I designed Portfolio Operating Intelligence around that review. The central product question is whether a shared view can improve the quality and speed of diligence while preserving the differences between portfolio companies.

Standardize the review, preserve company context

The portfolio view makes initiatives comparable through a common lifecycle and priority ranking. Opening an initiative adds its proposed value lever, impact and effort, vendor approach, implementation process, and 30/60/90-day plan. Company views add operating conditions and stakeholder context.

The tradeoff is deliberate: a common framework helps an operating team compare opportunities, but a uniform score cannot resolve differences in data readiness, management capacity, customer risk, or implementation cost. The ranking starts a discussion; the underlying evidence should determine the decision.

Separate reusable capabilities from local execution

Shared vendor views and cross-company playbooks make reuse visible. An operating team can examine whether contract abstraction, reporting, customer intelligence, or another capability has a common foundation across several businesses.

Reuse should depend on compatible workflows, data, and controls. A shared playbook can establish an approach and validation criteria while each company retains responsibility for integration, change management, and operating results. Central coordination should reduce repeated discovery without forcing every company into the same implementation.

Use AI to challenge the plan

The live advisor accepts questions in the context of the portfolio, a company, an initiative, or a playbook. That makes questions such as “What should we validate before scaling?” more useful than a generic conversation about AI strategy.

The product role of the advisor is to help frame risks, dependencies, and next questions. Its generated responses require review against source evidence. It does not approve an investment, verify a vendor claim, or establish that a proposed return will materialize.

What is implemented

The React interface connects a fictional portfolio of 11 companies and 33 initiatives with lifecycle views, company assessments, vendor relationships, example stakeholder digests, and eight playbooks. A live AI advisor uses the supplied scenario context through an AWS Lambda and API Gateway backend.

The portfolio data is bundled demonstration content. Readiness ratings and RICE scores are authored examples; the interface does not derive the RICE components. Digest screens illustrate stakeholder communication rather than scheduled delivery. Lifecycle states illustrate a review model rather than enforced approval gates.

Keeping the portfolio interface separate from the advisor allows the operating model to be explored without generating an AI response at every step. A production version would also need durable records, authorization, evidence provenance, and auditable decisions.

How I would validate business value

Start with a small set of initiatives and record the baseline: time spent preparing reviews, repeated vendor evaluations, time to a decision, and implementation follow-through. Test whether the shared view changes decisions and actions before treating usage as proof of value.

For each pilot, establish an owner, implementation and run costs, adoption assumptions, a quality threshold, and a decision date. Separate cash savings, incremental contribution, and released employee capacity; capacity only becomes financial value when there is a credible plan to use it.

Scale when the evidence supports the economics and the operating team can sustain the workflow. Revise or stop when adoption, quality, or delivery costs undermine the case. Those are the production decisions this demonstration is intended to make easier to discuss.