← Insights
Enterprise Operating Effectiveness
2025 · Q2
13 min read
DA-Note 036

The operating model implications of embedded AI.

Embedded AI is quietly redrawing the operating model. The commercial implications are largely under-discussed.

Thesis

The dominant enterprise AI conversation is about adoption. The more consequential conversation — about operating model, span of control, and the unit economics of knowledge work — is happening only in pockets. Boards that delay this conversation by 12 to 18 months will find themselves restructuring under pressure rather than by design.

I. The under-discussed shift

Most enterprise AI discussion concerns adoption: which functions, which use cases, which vendors. The harder conversation — what embedded AI does to the operating model — is happening in a small number of forward boardrooms and is largely absent from public discourse.

McKinsey's research on the economic potential of generative AI, and BCG's work on AI-enabled operating models, both point to the same structural conclusion: when AI is embedded into the workflow of knowledge work, the optimal span of control rises materially, the optimal layering of management compresses, and the unit economics of professional functions — legal, finance, advisory, engineering — shift permanently.

II. Three operating model questions

We advise CEOs and boards to commission explicit, named work on three questions in the next 12 months.

  • Span of control

    What is the right span of control for managers whose teams are AI-augmented? Existing spans were calibrated to a pre-AI assumption set and are now structurally too narrow.

  • Layering

    How many layers of management are required when individual contributor productivity rises 30 to 50 percent? In most enterprises, the answer is fewer than today's structure assumes.

  • Make-versus-buy

    Which capabilities, previously outsourced for cost reasons, are now economically rational to bring in-house — and which retained capabilities are now economically rational to outsource or automate?

III. The risk of delay

HBR's recent coverage of organisational redesign emphasises a pattern: enterprises that defer operating model decisions in the face of a structural shift end up making them under crisis conditions, with worse outcomes for both the institution and its people. The boards that commission this work in 2025 will redesign by design; the boards that defer will redesign under pressure in 2027.

IV. What disciplined boards are doing

The pattern we observe in the disciplined boards: a named executive — typically the COO or Chief Transformation Officer — is given a 6 to 9 month mandate to produce a defensible operating model thesis, supported by independent advisors and tested against the enterprise's specific economics. This is not a strategy exercise; it is a structural design exercise. It belongs at board level, not in an operating committee.

References
  • McKinsey & Company, The Economic Potential of Generative AI, 2023
  • BCG, AI at Scale: Building the AI-Enabled Operating Model, 2024
  • Harvard Business Review, Redesigning the Organisation for the AI Era, 2024
  • Gartner, Future of Work Research, 2024

Delta Advisory notes draw on the published research of Gartner, McKinsey & Company, BCG, and the Harvard Business Review, alongside engagement-level commercial intelligence from our own work. Notes are editorial and do not constitute investment, legal, or regulatory advice.

Next note
AI Governance
2025 · Q4

The institutional posture required for AI vendor governance.