Glossary · Governance
Workflow absorption
Workflow absorption measures whether AI has actually changed how work runs (steps redesigned, cycle time reduced, errors cut) as opposed to adoption, which only counts access such as seats and logins.
Workflow absorption measures whether AI has actually changed how work runs (steps redesigned, cycle time reduced, error rate moved) as opposed to adoption, which only counts access like seats and logins. MIT's 2025 finding that 95% of GenAI pilots showed no P&L impact describes high adoption, near-zero absorption.
- Adoption counts access: seats, logins, completions. Absorption counts changed work.
- Absorption metrics: cycle time, error/rework rate, steps redesigned, cost per successful task.
- MIT, July 2025: 95% of enterprise GenAI pilots showed no measurable P&L impact.
- High adoption with near-zero absorption is the exact shape of a failed pilot.
- You cannot measure absorption without baselining the workflow before you change it.
Also known as: absorption vs adoption, AI absorption
Workflow absorption measures whether AI has actually changed how work runs (steps redesigned, cycle time reduced, errors cut) as opposed to adoption, which only counts access.
Adoption asks "did people get the tool?" Absorption asks "did the work change?" Those are different questions, they are not correlated, and confusing them is how an organisation reports success while nothing improves.
How workflow absorption works
Absorption is measured on the workflow, not the person. You pick a workflow, baseline it before anything changes (cycle time per instance, rework rate, escalation rate, cost per successful outcome) then instrument the same numbers after the AI is in place and compare. The measurement is a delta on real numbers, which is why the pre-baseline is non-negotiable: without it you are comparing a measured present against a remembered past, and memory is generous.
Adoption metrics, by contrast, are easy to collect and easy to move: seats provisioned, licences activated, logins, completions, prompt counts. They rise when you run a training push and decay a few weeks later, and none of them tells you whether a single process runs faster. Absorption is harder to move and harder to fake, which is precisely what makes it worth reporting. Establishing and tracking it is the point of an evaluation harness applied to the workflow, not just the model.
Why workflow absorption matters for enterprise AI adoption
The distinction explains the headline failure of enterprise AI. MIT's NANDA study found that 95% of enterprise generative AI pilots produced no measurable P&L return, locating the cause in organisational workflow gaps rather than model quality (MIT NANDA, July 2025). Read through this lens, most of those pilots had adoption (seats were used), and near-zero absorption. The dashboard was green and the work was unchanged.
The figure has been contested as methodologically thin, and that criticism is fair; we make it ourselves. But the directional point survives the argument: organisations that measured access instead of change were surprised at renewal, and organisations that measured the workflow were not. For a transformation owner asked for a number this quarter, reporting an absorption delta on one workflow is a smaller, more defensible claim than a utilisation figure that will be used against them in a year.
Common mistakes with workflow absorption
The first mistake is never taking the baseline, then trying to prove improvement against memory. No pre-baseline, no absorption claim: only a story. The second is reporting adoption because it is easy and it goes up when pushed, while quietly hoping absorption followed. It usually did not.
The third is instrumenting the person instead of the workflow, which turns measurement into surveillance and poisons the honest reporting absorption depends on. Measure the process (cycle time, rework, escalation) at the aggregate level, and keep individual prompt logs out of the manager's hands.
Related terms
- AI governance framework: the controls that protect a real absorption gain.
- Evaluation harness: how the workflow delta is measured and held stable.
- Global capability centre: where absorption, not headcount, is the AI question.
- Agentic AI: the systems that only pay off if the workflow absorbs them.
How Chokmah approaches workflow absorption
Absorption, not adoption, is the number we sign up to move. It is our coined distinction and the spine of how we work. An adoption diagnostic starts by baselining the workflow before anyone touches a tool, and every engagement ends with a measured delta on cycle time, error rate or steps redesigned rather than a login count. We measure the workflow, not the logins, and we will report the number even when it comes back inconvenient, because a vendor who only measures what looks good is selling the thing that fails 95% of the time.
Sources
- MIT NANDA, The GenAI Divide: State of AI in Business 2025, July 2025. https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf
- AIGL, State of AI in Business 2025: figure breakdown, 2025. https://www.aigl.blog/state-of-ai-in-business-2025/
Related terms
- AI governance frameworkAn AI governance framework is the documented set of policies, roles, controls and records that determine who may deploy an AI system, on what data, with what testing, and who is accountable when it fails.
- Evaluation harnessAn agent evaluation harness is a repeatable test suite that scores an AI agent's outputs against fixed, versioned cases before and after every change, so teams can tell regression from variance.
- Global capability centre (GCC)A global capability centre is an offshore entity a multinational owns and operates to run strategic functions such as engineering, analytics, finance and operations in-house, rather than outsourcing them to a third party.
- Agentic AIAgentic AI is software that uses a language model to plan and carry out multi-step tasks by calling tools, observing the results, and choosing its next action in a loop.
Frequently asked questions
Adoption counts access: seats provisioned, licences activated, courses completed, prompts sent. Absorption counts change: workflow steps redesigned, cycle time reduced, error and rework rates moved. Adoption tells you a tool was opened; absorption tells you the work is different. They are not correlated: an organisation can have 90% of seats active and no process running any faster, which is a high-adoption, zero-absorption failure wearing a success dashboard.
Pick a workflow and baseline it before anyone touches a tool: cycle time per instance, rework rate, escalation rate, and cost per successful outcome. Then instrument the same numbers after the change and compare. The comparison only works if you took the pre-baseline: measuring afterwards against a remembered past is not measurement, because memory flatters. Absorption is a delta on real numbers, not a snapshot of activity.
Because absorption is the thing you were actually buying. Nobody funds an AI programme to raise login counts; they fund it to make work faster, cheaper or more accurate. MIT located the 95% pilot failure in organisational workflow gaps, not skills gaps: organisations measured adoption, saw it rise, and never checked whether the work changed. Absorption is harder to move and harder to fake, which is exactly why it is the number worth reporting.
It is the input ROI is built from. ROI is a financial ratio; absorption is the operational change underneath it: the reduced cycle time or error rate that, once costed, becomes the return. Reporting absorption first keeps the conversation honest, because it is measured on the workflow rather than modelled in a spreadsheet. A credible AI ROI claim traces back to an absorption delta someone baselined and can defend.
Put the concept to work
We install working agentic workflows, not vocabulary. Book a free AI Reality Check.