Build, buy or partner

How the capability is sourced, and what each route costs in control.

37documents on this topic
23organizations represented
4issues named
11sourced citations
0sourced statistics

The state of it

One of 4 topics within Operating model redesign.

The sourcing question has stopped being binary. 57% of organizations now prefer mixing built and bought agents rather than choosing one route - up from 51% a quarter earlier - and the direction is consistent across publishers: the analyst framing is build, buy or blend; the consultancy framing adds borrow, co-developing with a partner. What was a procurement decision is becoming a portfolio of them, made capability by capability.

The criteria the documents converge on are more useful than any verdict. Build where the capability is core differentiation, the data is too sensitive to leave, and deep orchestration control is required - and only with the engineering strength to sustain it. Buy where a vendor meets most of the need, differentiation is low, and dependency on the vendor's update cycle and decision logic is acceptable. Partner where internal capability is thin and risk-sharing matters more than ownership - accepting that what a partner supplies today commoditizes as agent marketplaces mature.

Two warnings sit under the choice. First, buying does not remove the work: the system around a model - the data it reaches, the context it is given, the integration into workflows - is still built, not procured, and that is where returns are decided. Second, the fully outsourced route quietly erodes the institutional knowledge that domain-heavy functions run on, a cost that never appears in the vendor comparison because it is paid years later.

The issues, by agreement

How many independent organizations name each issue as a problem. An issue is only as real as the number of separate publishers that identify it, so the count is the ranking. Bars are organizations, not documents. Where the count reads ours, no publisher here states the issue and the analysis is our own.

names it as a problemdisputes itqualifies it
The build-versus-buy binary is already obsolete
2
Buying the model does not buy the system
2
Fear of falling behind is making the sourcing decision
1
Full outsourcing quietly erodes the knowledge the function runs on
1

Who takes which position

The chart above counts positions; this shows whose they are. Read down a column for what one organization holds across the whole topic, and across a row for who lines up on one issue. Where a cell carries more than one position, the strongest is shown and the rest are in the tooltip.

Ddisputes it Qqualifies it Nnames it as a problem Pproposes a fix
Build, buy or partner: 4 issues against the 6 organizations cited on them. The number under each name is how many of these issues it is cited on.
Issue KPMG · 3 Gartner · 2 IDC · 2 EY · 1 IBM · 1 Microsoft · 1
The build-versus-buy binary is already obsolete N N P · · ·
Buying the model does not buy the system Q · · · N N
Fear of falling behind is making the sourcing decision N Q P · · ·
Full outsourcing quietly erodes the knowledge the function runs on · · · N · ·

A dot means this organization is not cited on that issue. It does not mean they are silent on it: an organization is cited where its document takes a position we could locate, and the absence of a citation is the absence of a finding, not a finding of absence. Who is represented lists everyone working on this topic, including those not cited above.

Where they disagree

No contradictions recorded on this topic yet.

The issues in full

Each issue carries the organizations that name it, the numbers behind it, and the remedies proposed - with the concrete steps under each. Every citation points at a section of a named document, so any count here can be checked.

Issue 012 organizations name it2026 evidence

The build-versus-buy binary is already obsolete

The decision is still framed as a single choice between building and buying. Most organizations have moved to a blend, and the choice is being made per capability, not once.

The blend is not indecision; it is the shape that matches how the capability actually arrives - a bought foundation extended with built components, or an internal agent augmented with specialized third-party modules. The practical consequence is that the organization needs a repeatable decision process rather than a one-time verdict: criteria, applied per capability, revisited as the vendor landscape shifts. A one-time verdict ages in months; the criteria do not.

How to fix it — 1 approach, 3 steps

Run the decision per capability, on written criteria

Replace the one-time verdict with a standing decision process: named criteria, applied to each capability, revisited on a cycle.

Done when Written criteria covering differentiation, data sensitivity, customization, lock-in and skills exist, every capability in the portfolio carries a recorded verdict against them, and revisits are triggered by vendor-landscape movement rather than the contract anniversary.

  1. Write the criteria once: differentiation, data sensitivity, customization, lock-in, skills.0-30 daysCIO
  2. Run the current portfolio through them and record the verdict per capability.30-90 daysChief AI officer
  3. Revisit verdicts when the vendor landscape shifts, not on the contract anniversary.ongoingCIO
The evidence — 3 documents
OrganizationDocumentPosition
GartnerConsultancy · March 2025Build, buy or blendOur reading Treats blended capability - foundation model interfaces combined with custom front-ends and integration - as a category alongside embedded and bought AI, reflecting how teams have shifted away from building models from scratch.Blended as its own categorynames it
KPMGConsultancy · January 2026Build-buy-borrow decisionOur reading Reports 57% of surveyed organizations preferring a mix of built and bought agents, against 51% the previous quarter, and describes blended approaches where an internal agent is extended with specialized third-party components.The blend is now the majority positionnames it
IDCConsultancyEnterprise build vs. buyOur reading Proposes anchoring every build-versus-buy assessment in the business outcome it serves, with the return case coordinated across technology, data science and line-of-business leaders rather than decided inside IT.Anchor the decision in outcomesproposes a fix

Issue 022 organizations name it1 qualifies it2026 evidence

Buying the model does not buy the system

A purchased solution is treated as delivered capability. The system that produces value - the data the model reaches, the context it is given, the workflow it sits inside - is still built in-house, and that is where the returns are decided.

The sharpest formulation comes from a convening of enterprise leaders: capability is a construction project, not a procurement one, and the organizations pulling ahead are not finding better models but building more deliberately around them. The purchase decision sets the floor; the construction around it sets the return. This also reframes vendor dependency: what is bought arrives with the vendor's update cycle and the vendor's decision logic, and the balance between internal and external capability determines whether the result is a differentiated engine or a weak, unoptimized assembly.

How to fix it — 1 approach, 3 steps

Budget the system around the purchase, not just the purchase

Every buy decision carries a construction line: integration, data preparation, workflow redesign and evaluation, estimated before the contract is signed.

Done when Every purchased capability has an integration and data estimate in its business case beside the licence cost, and realized construction cost is tracked against that estimate.

  1. For each purchased capability, estimate the integration and data work it still requires.0-30 daysCIO
  2. Put that construction estimate in the business case beside the licence cost.30-90 daysCFO
  3. Track realized construction cost against estimate, so the next case is honest.ongoingCFO
The evidence — 3 documents
OrganizationDocumentPosition
IBMHyperscalerEnterprise AI AgentsOur reading States that the balance between in-house and external capability determines whether the organization builds a powerful differentiated engine or a weaker, unoptimized solution, and recommends piloting both routes before committing.The balance decides the outcomenames it
MicrosoftHyperscaler · June 2026Tokenomics is the new headcountOur reading Argues that AI capability is increasingly a construction project rather than a procurement one - the model is the starting point, and the system built around it, the data it can access and the context it is given, is the work.Construction, not procurementnames it
KPMGConsultancy · January 2026Build-buy-borrow decisionOur reading Qualifies the bought route: ongoing dependency on the vendor for updates, support and the underlying model's decision logic, and limits on how far off-the-shelf solutions can be tailored to a specific domain.What the bought route carriesqualifies it

Issue 031 organization name it1 qualifies it2026 evidence

Fear of falling behind is making the sourcing decision

Solutions are bought because competitors announced theirs, not because a vendor met written criteria. The published decision frameworks exist precisely because the observed behavior is fear-driven.

One consultancy paper opens with the observation that fear of being left behind has driven off-the-shelf buying, then predicts the correction: more building and partnering as organizations mature. The antidote is not a slower process but a written one - the frameworks converge on the same short list of criteria: strategic differentiation, data and intellectual-property sensitivity, customization and orchestration control, governance requirements, integration depth, total cost and lock-in, and the skills available to sustain each route. A decision recorded against criteria can be revisited; a decision made from fear can only be repeated.

How to fix it — 1 approach, 3 steps

No sourcing decision without a written scorecard

Every material sourcing decision records its scores against the criteria before the contract is signed, so the reasoning survives the person who made it.

Done when The procurement template carries the shared criteria list, the last three material sourcing decisions have filed scorecards dated before their contracts, and a year of decisions has been audited against them.

  1. Adopt the shared criteria list and put it in the procurement template.0-30 daysCIO
  2. Score the next three sourcing decisions on it and file the scorecards.30-90 daysChief AI officer
  3. Audit a year of decisions against their scorecards: did fear or criteria decide.ongoingCFO
The evidence — 3 documents
OrganizationDocumentPosition
KPMGConsultancy · January 2026Build-buy-borrow decisionOur reading Observes that fear of being left behind led many companies to buy off-the-shelf solutions, and expects more building - in-house or with partners - as organizations mature, laying out a decision grid across strategic fit, data sensitivity, customization, governance, integration, cost and skills.Fear-driven buying, then correctionnames it
GartnerConsultancy · March 2025Build, buy or blendOur reading Locates the harm in the accumulation rather than in any single purchase - each departmental buy is treated as a reasonable answer to a real need, and it is the suite of them that produces overlap with features already owned, unnecessary cost and technical debt.The cumulative cost of unexamined buyingqualifies it
IDCConsultancyEnterprise build vs. buyOur reading Proposes assessing each decision across immediate and long-term cost, customization need, internal expertise versus hiring, time to market, security and compliance, vendor support and dependency risk, and strategic fit.The criteria listproposes a fix

Issue 041 organization name it2026 evidence

Full outsourcing quietly erodes the knowledge the function runs on

The transformation of a domain-heavy function is handed entirely to a provider. Delivery accelerates; the institutional knowledge of how the function actually works migrates to the vendor, and the bill for that arrives years later.

The admission is notable for where it appears: inside a managed-services provider's own paper, which concedes that the fully outsourced model risks eroding institutional knowledge and weakening alignment in high-context, domain-driven functions - and argues instead for a hybrid where the client retains ownership of decisions and insight while the provider supplies platform and delivery. The partner route has a second decay mode: what a partner supplies today commoditizes as agent ecosystems and marketplaces mature, so a partnership justified by exclusivity should be priced as temporary.

How to fix it — 1 approach, 3 steps

Name what must never leave before signing anything

Before any sourcing agreement, write down which decisions, data and domain knowledge remain in-house, and build capability transfer into the contract rather than hoping for it.

Done when A written list names the decisions, data and domain knowledge that remain in-house function by function, the contract carries capability-transfer milestones with dates and named recipients, and an annual review records what the organization could no longer do without the provider.

  1. List the decisions and knowledge that stay in-house, function by function.0-30 daysCOO
  2. Put capability transfer milestones in the contract, with dates and named recipients.30-90 daysGeneral counsel
  3. Review annually what the organization could no longer do without the provider.ongoingCOO
The evidence — 2 documents
OrganizationDocumentPosition
EYConsultancy · January 2026EY’s AI-first managed servicesOur reading Concedes that fully outsourced transformation risks eroding institutional knowledge and weakening alignment with core business priorities in high-context, domain-driven functions - an argument made in a managed-services provider's own paper.The provider's own warningnames it
EYConsultancy · January 2026EY’s AI-first managed servicesOur reading Proposes a hybrid in which the client retains ownership of key insights and decisions - defining model tolerances, approving automation thresholds - while the provider runs the platform, with capability transferred back as maturity grows.Retain decisions, source deliveryproposes a fix

Who is represented

This dossier is drawn from 23 organizations working on the subject, 6 of which are cited directly in the issues above.

Consultancy — 8

EY 2 KPMG 2 Gartner 1 IDC 1 Boston Consulting Group 6 McKinsey & Company 3 Capgemini 2 Cognizant 1

Institution — 2

Citi 1 World Economic Forum 1

Academic — 2

Carnegie Mellon SEI 1 National Bureau of Economic Research 1

Hyperscaler — 5

IBM 2 Microsoft 2 Google Cloud 3 AWS 1 Lenovo 1

Vendor — 6

IntuitionLabs 2 Andreessen Horowitz 1 Menlo Ventures 1 Predibase 1 WorkFusion 1 Writer 1