Concept card

Building The Human Foundation Of The AI-Powered Enterprise

As of 2026-10-10

Edited by Cédric Mary, sole editor and publication director — 27 years in the SAP ecosystem. How these pages are produced, and where automation is used.

What is Building The Human Foundation Of The AI-Powered Enterprise?

The binding constraint on AI-powered enterprises is the human foundation, not the model — Forrester predicts three in ten firms will harm their own growth in 2026 through AI self-service that never redesigned decision rights.

When to use this lens

When a client's AI programme is technically sound but adoption is stalling, the human-foundation lens is the diagnostic to reach for before commissioning another round of model tuning — it asks whether the organization, not the model, is the actual constraint.

What it is

In an April 2026 analysis, Forrester analyst R. W. argued that the binding constraint on AI-powered enterprises is not model quality, compute, or data quality — it is the human foundation: the roles, skills, incentives, and decision rights that have to be redesigned before humans and AI agents can actually work together. He organizes the prescription into three investment pillars — a human-led strategy that establishes purpose and intent before any tooling decision, human-focused operations that rebuild workflows and governance around augmented humans rather than around the software, and human-first transformation that funds change management at parity with the technology spend, not as an afterthought line item.

Why it matters

A beautifully architected Datasphere and Joule stack paired with a thin change-management workstream produces a six-month value-realisation lag that surprises only the steering committee.
When Joule drafts the FP&A variance commentary first, the analyst's job changes to certifying the agent's reasoning — if the role description isn't rewritten, the output sits unapproved in a queue.
Controllership, the AI governance council, and line management must agree in writing which decisions an agent can make autonomously versus which require human approval.

Key points

Forrester's three pillars: human-led strategy (intent before tooling), human-focused operations (workflows rebuilt around augmented humans), human-first transformation (change management at parity with technology spend).
Forrester prediction: 3 in 10 firms will harm their total experience growth in 2026 through poorly designed AI self-service — the failure mode is organisational, not technical.
Human-foundation workstream must be funded at parity with the platform workstream from day one, not bolted on at go-live.
Role redefinition before go-live: the FP&A analyst's new job is to interrogate and certify agent reasoning, not produce variance commentary — the role description must be rewritten and the analyst trained before the agent goes live.
Decision rights must be mapped in writing before the first autonomous agent action: which decisions are fully autonomous, which require human approval in the loop, which remain fully human.
Incentive alignment is structural: if a manager is measured on exception volume and an agent resolves exceptions, the manager's measured contribution disappears — incentives must shift to oversight quality and governance.
SAP Sapphire 2026 acknowledges 'serious change management' is required alongside the Autonomous Suite — the 50+ Joule Assistants each have defined KPIs that presuppose Warner's three pillars.
Diagnostic test: ask the sponsor to describe how the three most-affected roles change after go-live. If the answer is 'we haven't thought about that', the human foundation is not funded at parity.
Use when deploying agents into regulated decision chains (finance close, purchase-order release, payroll exceptions). Postpone when the organisation cannot articulate the operating model it is moving from.
The value-realisation lag in most SAP AI programmes is predictable and avoidable: it is caused by role ambiguity and missing decision-rights agreements, not by technical defects.

Terms used on this page

Human foundation: Roles, skills, incentives, decision rights enabling human-agent collaboration.
Decision rights: Formal allocation of who decides what when an agent is in the loop.
AI Council: Cross-functional body governing agent deployments in regulated processes.
SAP-managed Joule provisioning depth: Two levels for the SAP-managed Joule activation path: Level 2 (full customer-like formation, end-to-end testing) and Level 0 (rapid direct validation) — an operating-model choice disguised as a technical rollout option.
NVIDIA OpenShell: Open-source runtime SAP co-developed with NVIDIA that answers whether an agent action can safely execute — a technical safety check, not a business-authorisation decision.
Joule Studio runtime governance layer: The part of Joule Studio that answers whether an agent action should happen at all (business policy, identity, audit) — distinct from and not substitutable by NVIDIA OpenShell's safety check.
Oversight-quality KPI: A metric measuring the quality of a human's supervision of agent output, needed alongside (never instead of) volume-based KPIs so oversight is not structurally disincentivised.

Sources

Full card available to members. What the full card adds: the full decision framework · the SAP vs Snowflake / Databricks / Fabric comparison · the common pitfalls and their fix · the cheat sheet · the architecture schemas · the code blocks · the facts worth quoting.

Open in the app →