Build Compose Collaborate Buy — Four Paths for Enterprise AI Software

The framework

For decades, enterprise software strategy was a binary: build custom (expensive, slow, risky) or buy off-the-shelf (cheap, fast, generic). AI collapses the cost of the “build” side so dramatically that the binary dissolves into a spectrum of four options.

Path	What it means	Best when	Risk
Build	Construct proprietary systems atop foundation models, organized around specific organizational jobs	The system encodes distinctive data or decision logic that creates competitive advantage	Ongoing technical complexity; wasted if the workflow is not actually differentiating
Compose	Vendors provide scaffolding and primitives; business users configure for specific needs	The workflow matters but the underlying infrastructure does not; customization within guardrails is sufficient	Vendor constraints on how deep customization can go
Collaborate	Partner with forward-deployed engineers to build bespoke solutions in weeks, not months	Speed matters and the capability is not yet internalized; a bridge to eventual ownership	Dependency on external expertise; repeated reliance may mask a need to build internally
Buy Outcomes	Procure results rather than tools — “accurate financials as a service” instead of accounting software	The outcome is non-differentiating and can be metered objectively	Loss of control if competitive advantage later emerges in that domain

The governing decision

The four paths collapse into one diagnostic question: where does workflow ownership create competitive advantage?

Workflows that encode proprietary data, distinctive decision logic, or customer-facing differentiation should be built or composed.
Workflows that are operationally necessary but commodity should be bought as outcomes.
Workflows in transition — where the organization lacks capability today but may need it tomorrow — should be collaborated on as a bridge.

This maps onto a familiar strategy concept: the theory of the firm. Just as Coase asked “why do firms exist?” (because internal coordination is sometimes cheaper than market transactions), the AI-era version asks “which workflows should exist inside the firm?” The answer shifts as AI lowers the cost of both internal development and external outcome procurement.

What changes for incumbents

Enterprise AI spending jumped from $1.7B (2023) to $37B (2025). Over one-third of companies report replacing at least one SaaS tool with a custom AI alternative. SaaS valuations compressed 30-60% from 2021 peaks.

The implication for incumbent vendors: the “compose” and “buy outcomes” paths are where they survive. Salesforce Headless 360 is a compose-path adaptation. Adobe’s outcome-based CX Enterprise pricing (charging per successful AI-completed campaign) is a buy-outcomes adaptation. Vendors who insist on the old model — standardized tool, per-seat pricing, take-it-or-leave-it — lose the customers who now have cheaper alternatives.

Four implementation prerequisites

Nishar and Nohria identify four leadership decisions that determine whether the framework produces results or chaos:

Data architecture first. Fragmented, vendor-managed data constrains what AI systems can accomplish across all four paths.
Revisit build-vs-buy per function. The question shifts from “can it be built?” to “should it be built?” — and the answer differs by workflow.
Governance for internal systems. As development migrates outside traditional IT, security and maintainability become central concerns.
Technology as organizational design. When AI performs work previously handled by humans, technology decisions become workforce decisions.

Future of Software — parent topic tracking how AI reshapes the software industry
Seat-Based to Token-Based SaaS Pricing Transition — “Buy Outcomes” is the logical endpoint of the pricing shift from seats to results
Distribution as the Remaining Moat - Why SaaS Incumbents Aren’t Dead — counterpoint: incumbents with strong distribution survive even as the four-path framework pressures them
AI Transformation Requires Strong Form Org Redesign — “technology as organizational design” directly echoes the strong-form org redesign thesis
Run the Business vs Do the Work - The AI-Era Vertical SaaS Shift — the Build path maps to practitioners owning workflows; Buy Outcomes maps to outsourcing “run the business” functions
Absorb Automate Unbundle - Three Phases of Technology Deployment — temporal frame for how organizations move through these four paths over time
Hero User Strategy - Native AI’s Wedge Into Vertical Software — hero users are the ones who push organizations from “compose” toward “build”

Source

Deep Nishar and Nitin Nohria, “The End of One-Size-Fits-All Enterprise Software,” Harvard Business Review, April 23, 2026.