Agentic AI · Operations · Fractional leadershipMaking the impossible, possible.

AI has moved from answers to work: agentic AI

Two years ago, AI answered questions. Today agentic AI can draft the proposal, update the CRM, reconcile the invoices and flag the late shipment. It works through a task step by step as an agent, with a person approving the result. Typing a keyword and reading the answer uses a small fraction of what it can do.

The new org chart

In teams that have made the shift, each person leads a small crew of AI agents. The scheduler, the follow-up writer, the report builder. The person's job moves from doing every step to designing the work, checking it and owning the result. That is a leadership skill, and it can be learned.

How to pivot

1. Stop searching, start delegating. Give AI a whole task with context and an example, not a keyword.
2. Pick one task you dread. Hand AI the first draft of it every week for 30 days.
3. Write your process down. AI can only follow steps that exist on paper.
4. Learn to check the work. Facts, numbers and judgment calls stay with you.
5. Build one agent. A repeatable workflow for follow-up, reporting or research.
6. Measure the hours you save. Then show your team the number.
7. Teach someone else. It's the fastest way to become the person your team turns to for AI.

For leaders

Tools are the easy part. IBM's 2026 CEO study found that 83% of CEOs say AI success depends more on people adopting it than on the technology, yet only about a quarter of the workforce uses AI regularly. Train your people first, redraw the org chart around people leading agents, and measure the result.

Sources: Pew Research Center, U.S. workers and AI (Feb. 2025); Chatterji et al., How People Use ChatGPT, NBER Working Paper 34255 (Sept. 2025); Yahoo News on Richard Baldwin's 2023 remark; IBM Institute for Business Value CEO study, as reported by People Matters (May 2026).

1. Trust

Trust is built one kept promise at a time. Leaders who say what will happen and then make it happen, and who warn early when it won't, get the benefit of the doubt when things get hard. No model can borrow that history.

2. Judgment with incomplete information

Real decisions rarely come with clean data. AI can lay out options. A leader has to choose, explain why, and own the result.

3. Leadership and influence

Getting a team to change direction mid-course is a human job. It takes credibility, timing and knowing what each person needs to hear.

4. Solving messy, cross-team problems

The hardest problems sit between departments: sales promised it, operations can't build it, finance can't fund it. Untangling that takes someone who can see the whole chain and talk to every link.

5. Accountability

Strong operators tie every project to a number and a name. When the number moves the wrong way, they say so first.

6. Calm in a crisis

Resilience shows up when the plan breaks. The leader's job is to lower the temperature so people can think clearly.

7. Building and growing people

Teams that stay and improve are built by leaders who coach. That is still the best return on any hire.

How to spot these traits in an interview

• Ask for a time they missed a target and what they told their team.
• Ask how they made a decision when key facts were missing.
• Ask who they developed and where that person is now.
• Ask what they changed their mind about in the last year.

Where AI fits

Use AI to speed up the research, drafting and reporting. Then hire, and develop, leaders for the parts that need a person: trust, judgment and the will to see it through.

Source: World Economic Forum, The Future of Jobs Report 2025, Skills outlook.

1. A downtime log people actually use

A tablet at the machine, a short list of reason codes and a daily summary to the supervisor. Within weeks you know which three causes eat most of your hours.

2. Maintenance before the breakdown

Schedules based on hours run and past failures, with reminders that go to the right tech. Planned work is cheaper than emergency work.

3. Same-day quotes

Pull material, labor and machine costs from the spreadsheet or ERP you already have, and draft the quote automatically for a person to check and send. Faster quotes win more orders.

4. Shift handoffs that don't get lost

Digital handoff notes, searchable by machine and date, so the night shift doesn't repeat the day shift's problems.

5. Quality checks with photos

Checklists on a phone, photos attached, and an alert when a result is out of spec.

6. Training from your own SOPs

Short how-to videos and an AI assistant that answers questions using your procedures, so new hires get productive sooner.

The KPIs to watch

• Overall equipment effectiveness (OEE)
• Unplanned downtime hours by reason
• Scrap and rework rate
• Quote turnaround time
• On-time delivery and first-pass yield

Start with one line, one KPI and 30 days. Prove it, then roll it across the plant.

Source: The Manufacturer, reporting on The Manufacturing Institute and Deloitte workforce study (Sept. 2025).

Connect the whole flow

Lead, proposal, signed scope, project, time, invoice, payment. When those steps share one record, utilization and margin stop being a monthly surprise.

Dashboards partners will actually open

• Billable utilization by person and practice
• Realization and project margin
• Pipeline coverage for next quarter
• Proposal win rate by offer and source
• Days sales outstanding

Automations that give hours back

First-draft proposals from your past wins, kickoff checklists that run themselves, weekly client status reports built from project data, and client portals that answer the "where are we?" question before it's asked.

AI agents with a person in charge

Meeting notes turned into tasks, research briefs before a pitch, and first drafts of deliverables, all reviewed by the consultant who owns the client.

It scales both ways

A three-person boutique and a 300-person firm need the same logic at different sizes. Start with the biggest leak, usually pipeline or time capture, and build from there.

Source: Beancount, summarizing SPI Research's 2025 Professional Services Maturity Benchmark.

1. Define the end state and the date. What does done look like, and when?
2. Work backward. Find the critical path: the tasks that move the finish date if they slip.
3. Name the real constraint. People, money, time or materials. Plan around the tightest one.
4. One owner per line. Every workstream gets a single name, not a committee.
5. Set a battle rhythm. Short daily check-ins and a weekly status with the same format every time.
6. Build the dashboard first. If you can't see progress, you can't steer it.
7. Plan for friction. Write a Plan B for the top three risks before they happen.
8. Run an after-action review. What was supposed to happen, what did, why, and what we keep.

Most "impossible" projects fail on steps 3, 4 and 6. Fix those and the rest gets much easier.