When Small AI Teams Duplicate Investment

Your divisions are each running small AI-augmented product teams, and each team is delivering. Features are shipping, cycle times have improved, and the team leads are reporting progress sprint by sprint. These teams are small by design: focused mandates, fast feedback, tight loops between AI tools and human judgment. The question that no single team’s reporting can answer is whether the team one division over is solving the same customer problem.

In Brief


  • Setting the product definition as a funding condition, before releasing the first budget, distinguishes a portfolio that scales from one that rationalises.
  • Two teams sharing a product scope pay for tooling and infrastructure twice, with that overhead running undetected for over a year.
  • Investment committees review one business case at a time and have no current map of the customer problems existing funded teams are already covering.
  • Sprint-by-sprint progress reporting confirms execution fidelity; it cannot surface that a parallel team is funded to build the same product.

When small AI teams duplicate investment, the cause is rarely in their execution. The teams are not performing poorly. The investment governance that approved each initiative worked. What it did not establish, before releasing the first budget, is whether two separate funded backlogs were describing the same product.

Reporting hides duplication by design

A product team’s performance reporting is built to surface execution fidelity. When a team reports throughput, sprint delivery, and customer satisfaction metrics, it is accurately answering the question it was funded to answer: are we executing our mandate well? The question it has no mechanism to answer is whether a parallel team is executing the same mandate under a different budget line.

In the portfolio I document in Evolve, five of eleven AI-augmented investments were solving the same customer problem, funded and governed separately for fourteen months (Hodgson, 2026).

Each team was executing competently against the initiative brief it was given; that brief was the problem.

It was written from inside the initiative and reviewed against that initiative’s own terms. This constraint recurs across portfolios that have strong governance. Business cases for new AI team investments are written by the people proposing them. Those people know what they intend to build, and have limited knowledge of what every other funded team is already building. Investment committees review those cases one at a time, in the sequence they arrive. Neither party is poorly intentioned. Each is operating with an incomplete picture of what the full portfolio covers, because no one assembled that view before the first approval was made.

The Nexus Guide identifies the structural root: dependencies between teams come mainly from mismatches in product structure and communication structure (Schwaber, 2021). Coordination mechanisms, including shared planning, cross-team reviews, and dependency tracking, manage the symptoms of that original decision. They do not change what was funded.

Coordinating two teams that share a product scope does not reduce duplication. It adds overhead, including integration work, alignment meetings, and shared tooling decisions, to an investment that is already paying twice for the same customer problem. As I describe in Evolve, coordination is not accountability (Hodgson, 2026). Two teams with two ordered backlogs remain two teams with two ordered backlogs, regardless of how frequently they sync.

Investment reviews miss product overlap

Investment committees review each business case on its own terms: does the problem statement hold, does the proposed team have the right capability, does the investment thesis make sense? That review is thorough for the initiative in front of it. It cannot assess an initiative approved at a different time, by a different committee, under a different budget holder.

An investment committee approving a new team rarely has a complete picture of every other funded initiative at the product level. Business cases are prepared independently and reviewed in sequence. The executive reviewing a new AI-augmented team initiative has a well-prepared document describing the customer problem the new team will solve. What they lack is a current map of which customer problems the existing funded teams are already working on. Duplicate scope is invisible at the approval stage; the committee’s vantage point is a single well-prepared proposal, with no view across all currently funded products.

Each quarter both teams operate, the organisation pays more for the same customer problem. An organisation can carry separate tooling, AI infrastructure, onboarding, and organisational overhead across two teams for fourteen months before a portfolio review makes the overlap visible. By the time the rationalisation conversation is unavoidable, two functioning teams need to be reconciled. Their separate codebases, stakeholder relationships, and product definitions, built from different starting assumptions about what the product is for, all require resolution.

One backlog, set at funding

The Scrum Guide establishes the structural rule: multiple Scrum Teams on the same product share one Product Goal, one Product Backlog, and one Product Owner (Schwaber & Sutherland, 2020). That structural rule applies whether the teams use Scrum or not: a product is one set of users, one market condition, one ordered set of customer problems. Two funded backlogs describing the same users and the same problem set are describing one product, regardless of how they were approved.

The product definition — who the product is for, what problem it solves, and what a good outcome looks like — is a funding-time decision. I call this the funding rule: set the product scope before releasing the first budget. Setting it after teams are already running is expensive. It requires the kind of reconciliation work that portfolio rationalisations are built from.

Before approving a new AI-augmented team, the investment governance question is whether the proposed product scope overlaps with any existing funded initiative. Where the answer is yes, the decision is whether the existing product team has the right product owner, the right backlog order, and the capacity to absorb the additional work. Scaling down can deliver more value than adding teams (Schwaber, 2021).

With a single Product Owner and an ordered backlog, adding teams scales throughput; without one, it scales overhead instead.

The COO who establishes the product definition as a funding condition, before approving the next round of teams, makes each investment legible against what already exists. Establishing the product definition before the first budget is released is what distinguishes a portfolio that scales from one that rationalises.

References

  1. Hodgson, M. (2026). Evolve: The operating model AI demands. Zen Ex Machina. ISBN 978-1-7649380-1-3
  2. Schwaber, K. (2021). The Nexus guide. Scrum.org. https://www.scrum.org/resources/online-nexus-guide
  3. Schwaber, K., & Sutherland, J. (2020). The Scrum guide: The definitive guide to Scrum: The rules of the game. Scrum.org. https://scrumguides.org/docs/scrumguide/v2020/2020-Scrum-Guide-US.pdf

About the author

Receive insights on strategy, leadership, and transformation.
By subscribing you agree to our Privacy Policy
© 2026 Zen Ex Machina (ZXM) Pty Ltd. All rights reserved. ABN 93 153 194 220
search previous next tag category expand menu location phone mail time cart zoom edit close