The Agentic Workforce Assessment: What It Is, What It Costs, and What It Tells You

An agentic workforce assessment is a structured analysis of how work actually gets done in your organization: which tasks AI agents can take on today, which tasks need a human in the loop, which decisions must stay human entirely, and how much capacity moves when you act on that map. The deliverable is a board-ready answer to a simple question: where, specifically, is the opportunity in our company, and what is it worth?

I want to define the term carefully because the phrase is new and the space around it is filling up with things that are not assessments. Some are strategy essays from the big consulting firms. Some are free quizzes from software vendors that exist to qualify you as a sales lead. Neither tells you where the hours are in your business. This piece is my attempt to define the category honestly, including what the work costs and what you should demand from anyone who offers it.

Start with why this matters right now. According to McKinsey's 2025 State of AI survey — a self-selected sample of roughly 2,000 executives that skews toward large organizations — 88% of respondents said their organizations regularly use AI in at least one business function. But the same survey found nearly two-thirds had not begun scaling AI across the enterprise, and only about 7% reported AI fully scaled. The adoption story is over. The scaling story has barely started, and the gap between the two is where the money is.

The agent-specific numbers are starker. As of McKinsey's mid-2025 fieldwork, 62% of respondents said their organizations were at least experimenting with AI agents — but in any single business function, no more than 10% said they were scaling them. Translation: nearly everyone is running experiments, and almost nobody has done the organizational work to turn experiments into operating capacity. That organizational work starts with knowing your own task structure, which is exactly what an assessment produces.

There is also a failure wave coming, and you should know about it before you spend anything. Gartner forecast in June 2025 that over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. That is an analyst forecast, not a measurement — but the three causes it names are precisely the things an honest assessment is designed to settle before you commit budget: what it costs, what it is worth, and how it gets governed.

So what is the work, concretely? At SDI Clarity the assessment runs five weeks and has four parts. Part one is process mapping. We sit with the people who do the work and map the workflows as they actually run — not as the org chart or the process documentation says they run. Those two things are never the same, and the difference is usually where the opportunity hides.

Part two is task segmentation. We break roles into tasks and sort each task into three buckets: work an agent can own today, work an agent can assist but a person must finish, and work that stays human because it carries judgment, relationships, or accountability. This is the heart of the method. Agents do not replace jobs; they absorb tasks. If your analysis is at the job level, it is wrong before it starts.

Part three is opportunity sizing. We put hours and loaded cost against each task bucket and produce a capacity number — how many hours move, in which roles, worth how much annually. The assumptions are conservative and stated, so your CFO can rerun the math and challenge it. I would rather hand you a defensible number than an impressive one.

Part four is the roadmap: a sequenced plan with owners, the first ninety days spelled out, and the governance questions answered — who sets an agent's decision boundaries, who monitors it, what gets audited. Only 21% of respondents in Deloitte's 2026 State of AI in the Enterprise survey (3,235 business and IT leaders, director level and up) said their organizations have a mature governance model for agentic AI. The roadmap exists so you are not in the other 79% when something goes wrong.

What does it cost? It starts at $25,000 and is tiered by organization size: $25,000, $50,000, or $75,000, published. Five weeks. The big firms quote six figures and six to twelve months for the equivalent diagnostic, and the work is typically done by junior consultants and delivered as a strategy deck. I publish the tiers because I would rather have a fact-based conversation than a scoping dance, and because a mid-market company should be able to know what a decision costs before making it.

Who is it for? Organizations in the 500-to-2,500-employee range. Below that, the capacity math rarely justifies a formal engagement — start with the free resources on this site instead. Above that, you have transformation offices and enterprise vendors competing for your attention already. The mid-market is the band where the opportunity is real and the honest help is scarce. McKinsey's own 2025 survey shows the size gradient: 29% of respondents from companies under $100 million in revenue had reached AI scaling, versus nearly half from companies above $5 billion. Smaller organizations are behind on scaling even among the engaged — which means the capacity is still on the table.

A word on who does the work, because I think it is the most important question in this category. I run AI agents in my own companies every single day. I test what they can do every morning, I watch where they break, and I redesign my own workflows around what I learn. That is the evidence base I bring into an assessment. The large firms publish surveys of other people's work. A survey tells you what a thousand executives said. An operator tells you what the tools actually did. I tend to trust real people who are actually using these systems — I do not trust marketing speak by default, and you should not either.

I will also be straight about the value question, because the honest data is messier than either the optimists or the skeptics admit. MIT NANDA's July 2025 report — non-peer-reviewed, and contested — reported that roughly 95% of custom enterprise GenAI pilots showed little or no measurable P&L impact within about six months. Meanwhile PwC's May 2025 AI Agent Survey (308 US senior executives, self-reported) found 66% of adopters reporting measurable productivity value. Both numbers are out there, and they cannot both be the whole story. My read: individual productivity gains are real and widespread, but P&L-level impact is rare because it requires organizational redesign, not tool purchases. The assessment exists to find where the impact is real in your organization before you spend like it is real everywhere.

Now the section I would want if I were the buyer: questions to ask any assessment vendor, including me. First — who actually does the work? If the partner sells the engagement and analysts you never meet deliver it, you are buying a brand, not an assessment. Second — does the assessor run agents themselves, in their own operation, or do they only write about other people's deployments? Ask for specifics. The answer is revealing.

Third — is the assessment tied to a software platform? If the company assessing you also sells the agents, the assessment is a sales instrument. It may still be useful input, but it will find that you need their product. Every time. Fourth — is the price published? An unpublished price means the price is whatever the scoping call decides you can pay. Fifth — and this is the one that separates the honest vendors from the rest: will they tell you where agents will NOT work in your business? Ask them to name the parts of your operation they would leave alone. If every answer is yes, walk away.

Sixth — ask how they handle the numbers they quote. If a vendor tells you "88% of businesses use AI," they are misquoting an enterprise survey as a population fact; representative US Census data puts overall AI use at roughly one in five US businesses. A vendor who is careless with other people's statistics will be careless with your capacity math.

What you walk away with, at the end of five weeks: a map of where your workforce capacity actually is, the math behind it stated conservatively enough to defend in a board meeting, a sequenced roadmap with owners and governance, and — just as valuable — a clear list of what not to do. Most assessment buyers are surprised by how much of the value is in the second list.

My operating philosophy, in one line: I would rather be 80% right and acting in the real world than 100% perfect in theory. The companies that win the next five years will not be the ones with the most polished AI strategy deck. They will be the ones that found their real opportunities early, moved on them, measured what happened, and corrected fast. An assessment is how you start that loop with your eyes open.

If you want to go deeper on specific pieces of this, I have written companion articles on justifying AI investment to your board, on what agents can and cannot do today, and on how to measure the capacity they create — all here on the blog. And if you want the conversation about your own organization, the Agentic Workforce Assessment page on this site has the full methodology, timeline, and a direct line to my calendar.