Why you need an Agentic Clinic: Identify use cases and scale like a real enterprise
How a 2-day consultative format took a global aerospace manufacturer from 20 fragmented systems to scaled agentic AI in just 14 weeks.
In April, cross-functional leaders from a UK-based global aerospace manufacturer's sites across the UK, Ireland, and the Netherlands spent two days with Ema in London.
By late July, they had identified seven agentic AI use cases, a scoped 2-country pilot, and created a live HR AI employee that was tested across multiple policies, with 86% of rated responses scoring favourably. They also had an approved API integration with the company's HR system of record for a 400-user MVP, enabling Ema’s HR AI employee to take real HR actions.
These were the first steps on a scaling path toward a 16,000-person global workforce.
The starting conditions made that pace remarkable. Decades of growth through acquisition had built a business of 16,000 employees across more than 30 manufacturing sites in 12 countries,.It was a deeply fragmented estate: at one point more than 20 HR management systems, nine ERPs mid-way through a multi-year consolidation, and strict aerospace-grade data-sovereignty walls between regions. The cost landed on people: 1,600 hours a year of onboarding administration in the Netherlands alone, a thousand supplier queries a month hitting one shared-service team, shop-floor workers burning thirty-minute lunch breaks on password resets. The Agentic AI Clinic is the format Ema built for exactly this situation. Here is how it works, and why the outcomes follow.
The Shallow Dive: Information Gathering
In the run-up to London, each of the manufacturer's workstream leads completed a structured Shallow Dive with Ema: a short, focused session that mapped the function's processes, systems and headline volumes without attempting to solve anything.
It arms the facilitators with each function's vocabulary, system landscape and organisational context, so questioning starts at the specifics.
And it identifies the data and documents each lead should bring, which is how Ema's team knew exactly where the policy repository lived when the moment came to build an agent live.
The Deep Dive: When Breadth Becomes Depth
Then follows a two-day clinic in London, which is the Deep Dive. Where the Shallow Dive maps each function's landscape at a high level, the Deep Dive is where the organisation goes below the surface.
Processes are decomposed step by step, volumes are challenged and quantified in the room, use cases are prioritised against value and feasibility, and the platform is proven live against the customer's own documents.
The Deep Dive runs on three disciplines, and the rest of this piece walks through each:
- teach, because the customer's frame usually needs challenging before their processes do;
- tailor, because value must be argued in each stakeholder's own numbers;
- and prove, because nothing builds conviction like watching an agent stand up on your own data. Two days is enough, but only because the Shallow Dives made every hour count.
Teach, Don't Pitch: The Challenger Move That Reframed the Problem
The clinic runs on a challenger-sale conviction. Customers don't need another vendor agreeing with their roadmap; they need a partner willing to reframe it. The manufacturer's instinct, like most enterprises mid-consolidation, was that intelligent automation had to wait for the five-year ERP standardisation to finish.
But Ema's team believes in a different premise on Day 1: Agentic AI should be a layer over the systems you already own, not a one-to-one replacement or addition to them.
A Generative Workflow Engine orchestrating reusable agents across the document estate, the HRIS and nine ERPs means the work that hurts today gets automated today, with no rip-and-replace and no waiting for the estate to be tidy.
Although procurement wasn't on the original agenda, we pushed to include it so that 1500 suppliers could start being paid on time.
Constructive tension was welcomed rather than smoothed over. The company's leadership said they wanted several small, high-ROI successes over six months rather than a three-year strategic roadmap, and that IT capacity, not ambition, was their scarcest resource. So we negotiated the operating model in the room: fast proofs-of-concept where validation is the goal, fuller pilots where integration effort is justified by value, and a delivery design built on Ema's managed cloud and no-code tooling to minimise IT dependency.
Listen Like a Consultant, Build Like an Engineer
If the challenger posture supplies the reframe, the consultative discipline supplies the evidence. Because the Shallow Dives had already established each function's landscape, the room skipped generic discovery and interrogated specifics. Function by function, the customer's own people put their processes on the table with unusual candour: HR quantifying 800 onboarding flows a year at two hours of administration each; the finding that two-thirds of shop-floor queries are time-and-attendance questions asked by workers with no corporate email or device, which redirected the design toward voice and kiosk channels; IT mapping password resets and a folder-access problem rooted in years of data sprawl; Finance sizing its supplier helpdesk and its manual reporting cycle. Every use case left the room with an owner, a volume, and an ROI hypothesis, the raw material of a business case, not a wish list.
Consultative listening earned the right to the clinic's decisive moment: a live build. Using Ema's Autopilot, the team created an HR policy Q&A agent in minutes, grounded in the manufacturer's own policy documents pulled from its document management system, answering real questions with cited sources while the room probed it on guardrails, simplified answers and next-step actions. One senior stakeholder asked for a recording to take into a business review the following Monday. Tailoring the message to each stakeholder, with inventory reduction for supply chain, deflection volumes for IT, and employee experience for HR, is what turned one workshop into seven sponsored use cases.
The Only Metric That Matters:
Within a week of the Deep Dive, follow-up scoping sessions ran for each prioritised workstream, validating volumes, mapping systems and stress-testing every ROI hypothesis.
By late May that discipline had converted the prioritisation grid into the costed portfolio the teams called the Super 7, and the Super 7 into a scoped, dual-track pilot: an HR AI employee for a ~400-person UK engineering and technology centre handling policy Q&A plus first actions such as leave applications, and a Finance supplier-query assistant in the Netherlands. The architecture carried the clinic's fingerprints: two independent Ema instances, UK and Netherlands, designed around the data-residency constraints the customer's team had laid out on day two.
The Takeaway: Outcomes Are a Method, Not a Hope
Enterprises rarely lack imagination about agentic AI. T; they lack a defensible answer to where it should start.
This manufacturer's 14 weeks shows what it takes to close that gap:
Shallow Dives to qualify the room, and a 2-day Deep Dive that teaches a new framework instead of pitching features, quantifying every claim through consultative listening. Weprove the platform live before disciplined scoping converts momentum into a signed statement of work, resulting in seven costed use cases, a signed pilot, an 86% user-validated AI employee, a production integration path, and an expanding map of what comes next for 16,000 employees worldwide.
If your organisation is weighing where agentic AI fits across HR, IT, Finance or the supply chain, an Agentic AI Clinic is the fastest route from speculation to a plan your board can measure. Talk to us about running one for your team.