Automation discovery workshop agendas look almost too simple from the outside: a room, a whiteboard, and a day of your best operational people’s time. What happens in that room, though, decides whether the next twelve months of automation spending produces a saving somebody can point at in a budget line, or a shelf of licences nobody opens.
This guide is about what to expect. Not the marketing version, where everyone leaves inspired, but the operational version: who should be sitting there, what you must send them beforehand, what the day feels like hour by hour, what you should refuse to leave without, and the warning signs that the session is really a sales meeting wearing a facilitator’s badge. It is written for mid-sized UK organisations commissioning their first serious piece of business process automation, whether the facilitation is bought in or run in-house.
A well-run automation discovery workshop is a decision-making exercise, not an ideas session. It exists to turn a vague ambition — “we should automate more” — into a ranked, costed shortlist with a named owner against the top item. Our business process automation, workflow automation and intelligent automation engagements almost always start here, because the alternative — choosing by whoever complains loudest — is reliably expensive.
If you already know which process you want to tackle, pair this with our guide to choosing the first process to automate and the automation ROI calculator. If you are still weighing tool categories, read RPA vs workflow automation vs AI agents first.
Table of contents
- What an automation discovery workshop actually is
- Who belongs in the room for an automation discovery workshop
- How to prepare before the automation discovery workshop
- The seven stages of an automation discovery workshop agenda
- What to expect hour by hour in an automation discovery workshop
- The deliverables an automation discovery workshop must produce
- What an automation discovery workshop costs and who should pay
- Red flags: when an automation discovery workshop is really a sales meeting
- Remote, hybrid or in person: choosing the automation discovery workshop format
- The thirty days after your automation discovery workshop
- Mistakes that waste an automation discovery workshop
- Measuring whether the automation discovery workshop worked
- Frequently asked questions about the automation discovery workshop
- References
What an automation discovery workshop actually is
The name does a lot of work and most of it is misleading. “Workshop” suggests collaboration and sticky notes; “discovery” suggests exploration. Both are half right.
The one-line definition
An automation discovery workshop is a facilitated session in which a small group examines a pre-assembled list of candidate processes, quantifies each one against agreed criteria, and leaves with a ranked shortlist and a decision about what happens next. Everything else that happens is decoration.
It is a decision-making exercise, not an ideas session
The most common failure mode is treating the day as a brainstorm. Ideas are not scarce in any organisation — every team already knows three things they would like automated. What is scarce is the evidence to rank them and the authority to choose. An automation discovery workshop that generates forty new suggestions and ranks none of them has actively made your position worse.
How it differs from a vendor demonstration
A demonstration shows you what a product does and then looks for a process that fits it. Discovery starts from your processes and only reaches technology at the very end. If tooling comes up before lunch, the day has been inverted.
Where it sits in the wider programme
Discovery sits after the strategic decision to invest and before any procurement. It is the bridge between “we have budget for automation” and “we are building this specific thing”, and it should cost a fraction of one per cent of the programme it governs.
| Dimension | Discovery workshop | Idea brainstorm | Vendor demo |
|---|---|---|---|
| Starting point | Pre-built candidate longlist | Blank whiteboard | The product |
| Input required | Volume and handling-time data | Opinions | A meeting room |
| Attendees | 6 to 10, hand-picked | Whole department | Whoever is free |
| Output | Ranked, scored shortlist | Unranked idea list | A proposal |
| Success test | A decision was made | People felt heard | A follow-up was booked |
| Technology discussed | Last hour, if at all | Constantly | Only technology |
| Typical duration | One full day | 90 minutes | One hour |
Who belongs in the room for an automation discovery workshop
Attendance is the single strongest predictor of whether the day produces anything. Six to ten people is the working range; beyond twelve the session becomes a presentation.
The process owner
Somebody must be able to say “yes, those are the rules, and I can change them”. Without that person the workshop can describe a process but cannot commit to automating it, and every conclusion becomes provisional.
The people who actually do the work
Two or three practitioners, chosen because they handle the volume rather than because they manage it. They are the only reliable source of exception rates, and they will contradict the documented procedure within the first hour. That contradiction is the most valuable thing the day produces.
A systems or integration person
Someone who knows what the finance system will and will not expose, whether there is an API, and which integrations are already held together with a scheduled export. Their job is to convert enthusiasm into effort estimates before anyone commits.
A finance representative
Not to approve spending, but to define what counts as a benefit. A saving that never leaves a cost centre is not a saving, and it is far cheaper to learn that during the automation discovery workshop than in month nine.
The sponsor, for the first and last thirty minutes
The sponsor opens the session by stating the constraint — budget, deadline, or the problem that triggered it — and returns at the end to accept or reject the shortlist. Sponsors who stay all day tend to compress the disagreement that the session depends on.
Who should not be there
Anyone attending to defend a territory, anyone from a vendor whose product is a candidate, and anyone invited purely so they do not feel excluded. Political attendance is the most reliable way to convert a working session into theatre.
| Role | What only they can supply | Time needed |
|---|---|---|
| Process owner | Authority to change the rules | Full day |
| Practitioners (2 to 3) | Real exception rates and workarounds | Full day |
| Systems or integration lead | Feasibility and interface reality | Full day |
| Finance representative | Definition of a countable benefit | Afternoon |
| Data or reporting analyst | Volume figures from source systems | Morning |
| Risk, legal or DPO | Personal-data and decision-making limits | One hour |
| Sponsor | The constraint, and the final acceptance | First and last 30 minutes |
| Facilitator | Neutrality and timekeeping | Full day |
How to prepare before the automation discovery workshop
Preparation is where most of the value is created. A day with good inputs finishes early; a day without them becomes a data-gathering meeting that ends in a request for another meeting.
Pull the volume data first
Counts of cases per month, per queue, from the source system rather than from memory. People routinely misestimate their own volumes by a factor of three, and every downstream calculation inherits that error. Two hours of query writing beforehand is worth more than any facilitation technique.
Collect artefacts, not opinions
Gather the actual objects the process runs on: the spreadsheet, the form, the email template, three real historic cases including one that went wrong. Discussion anchored to a real artefact converges; discussion anchored to recollection does not.
Circulate a longlist in advance
Send the candidate list two weeks ahead with a line on each explaining why it is there. This turns the first hour from generation into confirmation, and it lets people arrive with objections already formed. Modelling any complex candidate in BPMN 2.0 notation beforehand is time well spent.
Agree the decision rights up front
Write down who decides, on what evidence, and by when — before the automation discovery workshop begins. Ambiguity here is what produces sessions that end in “we’ll take it away and think about it”, which is the polite form of nothing happening.
Book the right length
One full day for a first session with a longlist under twenty-five candidates. Two days if the process crosses three or more departments. Half a day is enough only when the shortlist is already down to two options and you are choosing between them.
The seven stages of an automation discovery workshop agenda
Every effective automation discovery workshop we have run or reviewed follows roughly the same seven-stage shape, regardless of sector. The proportions shift; the order does not.
Stage one: frame the objective and the constraint
Thirty minutes in which the sponsor states what the money is for and what the limits are. Cost reduction, capacity release, error reduction and compliance are different objectives that rank the same candidates differently, and picking one is not optional.
Stage two: confirm the longlist
Forty-five minutes to accept, reject or add candidates. Additions are welcome; the rule is that a new candidate must arrive with at least an order-of-magnitude volume estimate or it waits for the next round.
Stage three: walk the top processes end to end
Two hours, and the heart of the day. Take the five strongest candidates and walk each one from trigger to outcome, naming every system touched and every decision a human makes. This is where the documented procedure and the real one separate.
Stage four: quantify volume, handling time and error rate
Ninety minutes converting the walkthroughs into three numbers per candidate. Volume from the system, handling time from observation, rework rate from the last hundred cases. Estimates are fine; unsourced estimates are not.
Stage five: apply the disqualifiers
Thirty ruthless minutes. Any candidate whose process is mid-change, whose exception rate exceeds a third, whose source system is being replaced, or which has no willing owner leaves the list now regardless of its score.
Stage six: score and rank
An hour of structured scoring against agreed weights, done openly so that the disagreement surfaces in the room. The scores matter less than the argument they provoke; a candidate that everyone scores identically usually has not been understood.
Stage seven: agree the next step and the owner
The final thirty minutes. One process, one named owner, one date, one defined next artefact. An automation discovery workshop that ends without all four has not finished.
What to expect hour by hour in an automation discovery workshop
Knowing the emotional shape of the day helps enormously, because the middle of it feels like failure and is not.
The first hour is slower than you expect
Framing feels like preamble and people fidget. Resist compressing it — almost every unproductive session we have reviewed skipped the constraint-setting and spent the afternoon rediscovering that nobody agreed what “benefit” meant.
Mid-morning brings the first real disagreement
During the walkthroughs, two people who have worked alongside each other for years discover they do the same step differently. This is the point at which the room starts producing genuine information, and a good facilitator slows down here rather than moving on.
After lunch the numbers deflate the favourite
The process everybody wanted to automate turns out to run 300 times a year rather than 3,000, and the unglamorous reconciliation task nobody mentioned turns out to consume eleven hours a week. This reversal happens in roughly half the sessions we see and it is the clearest sign the day is working.
Late afternoon is where authority matters
Scoring produces a tie, or the top-ranked item belongs to the person with the least influence. Without pre-agreed decision rights this is exactly where an automation discovery workshop stalls and becomes another meeting.
The last thirty minutes decide everything
Read the commitment back out loud: the process, the owner, the date, the artefact. If anybody hedges, deal with it now. Hedged commitments do not survive contact with the following Monday.
The deliverables an automation discovery workshop must produce
Judge the session on its artefacts, not on the mood in the room at five o’clock. Five things should exist within three working days.
A scored and ranked shortlist
Three to five candidates with visible scores against visible criteria, so that a person who was not present can follow the reasoning. A ranking without the working shown is an opinion with a number attached.
An end-to-end process map for the top candidate
Trigger, steps, systems, decision points, exceptions and outcome. It does not need to be beautiful. It needs to be agreed by the people who actually run the process, ideally by initialling it before they leave.
A sized benefit case
Volume multiplied by handling time, adjusted for the proportion that genuinely automates, expressed in hours first and money second. Anything discounted over more than a year should follow the appraisal logic in the HM Treasury Green Book.
A feasibility and risk note
One page covering interfaces, authentication, data quality, personal-data exposure and blast radius. Where the process makes decisions about people, the note must reference a data protection impact assessment and the limits on automated decision-making.
A named owner and a dated next step
The shortest deliverable and the one most often missing. Without it, the other four become reference documents rather than a plan.
What is not a deliverable
A photograph of a covered whiteboard. A slide deck summarising the discussion. A proposal from the facilitator’s employer. None of these are outputs of an automation discovery workshop; they are souvenirs of one.
| Deliverable | What good looks like | Red flag |
|---|---|---|
| Ranked shortlist | 3 to 5 items, scores and weights visible | A single “obvious” winner |
| Process map | Every system and exception named | Four boxes and an arrow |
| Benefit case | Hours first, money second, sourced | A percentage with no denominator |
| Feasibility note | Interfaces and access confirmed | “Integration is straightforward” |
| Risk note | Blast radius and reversibility stated | Risk section left blank |
| Owner and date | One name, one date, in writing | “The team will pick this up” |
| Turnaround | Circulated within 3 working days | Arrives with the proposal |
What an automation discovery workshop costs and who should pay
Cost is where buyers get anxious unnecessarily, because the invoice is rarely the expensive part.
The internal cost is the real cost
Eight people for a day, plus perhaps twenty hours of preparation, is roughly two person-weeks of your own capacity. At mid-market UK rates that is a meaningful number before any external fee, and it is the number to protect by preparing properly.
Typical external facilitation fees
Independent facilitation of a one-day automation discovery workshop in the UK generally runs from a low four-figure sum for a single process area to the mid four figures for a cross-departmental session with a written report. Two-day formats and detailed process mining push it higher.
Free workshops and what they actually buy
Vendors and integrators frequently offer discovery at no charge. This can be genuinely useful, and it is never free: you are paying with your data, your people’s time and a strong presumption in favour of that vendor’s platform. Accept it with your eyes open, and never let it be the only assessment you commission.
How to judge value for money
The test is whether the output would survive a challenge from your finance director. A session that cost four thousand pounds and produced a defensible business case is cheap; a free session that produced a proposal is not, because the proposal costs you the next eighteen months if it is wrong.
Red flags: when an automation discovery workshop is really a sales meeting
Not every session badged as discovery is one. Five signals separate the two reliably, and all of them are visible before you commit.
The agenda arrives with a product name on it
A genuine agenda names stages and outcomes. If a platform appears in the running order, the conclusion has been written and the day is a route to it.
Nobody asks for your volume data
An automation discovery workshop that requires no preparation from you cannot possibly quantify anything. The absence of a data request is the clearest single tell.
The output is a proposal rather than a shortlist
Ask, in writing and in advance, what artefacts you will receive. If the answer is a statement of work, you are buying a sales process. If the answer is a scored shortlist you own outright, you are buying discovery.
Discovery is “free, and the fee is waived if you proceed”
This is a perfectly legitimate commercial offer, and it is also a conflict of interest written into the pricing. If the assessment only pays for itself when you buy, the assessment is not neutral.
The facilitator never says no
A good facilitator will disqualify candidates, including attractive ones, and will occasionally conclude that nothing on the list is worth automating yet. A facilitator who finds every idea promising is not assessing anything.
Remote, hybrid or in person: choosing the automation discovery workshop format
Format matters more than most buyers expect, chiefly because of what it does to the walkthrough stage.
In person, one full day
The default, and the strongest option for a first session. Walking a process end to end with six people around one table produces the contradictions that make the day worthwhile, and those contradictions surface far less readily on a video call.
Remote, three half-days
A good second choice for distributed teams, and better than a single long remote day, which nobody survives attentively. Split it as framing and longlist, walkthroughs, then quantification and scoring, with homework between each.
Hybrid, and why it usually disappoints
Four people in a room and four on a screen consistently produces a two-tier session in which the remote attendees stop contributing by mid-morning. If some attendance must be remote, make everybody remote.
What the format changes and what it does not
Format changes energy, pace and how much preparation you need. It does not change the seven stages, the deliverables, or the requirement for a named owner at the end. Judge any automation discovery workshop by its artefacts, not by whether it happened in a room.
The thirty days after your automation discovery workshop
The session is the cheap part. What follows determines whether it mattered, and the window is short — attention decays fast.
Week one: circulate and correct
Send every artefact to every attendee and invite corrections with a deadline. Practitioners routinely remember an exception two days later, and that exception is often the thing that would have derailed the build.
Week two: validate the numbers against history
Take the top candidate’s volume and exception assumptions and test them against the last hundred real cases. Roughly a third of the sizing estimates from an automation discovery workshop move materially at this step, which is exactly why the step exists.
Week three: prove the hard technical assumption
Every candidate has one assumption that carries the risk: an API that may not exist, a permission nobody can grant, a document format that varies more than anyone admits. Prove or disprove that single assumption before committing budget.
Week four: the go or no-go
Take the corrected case to the sponsor with a recommendation, not a menu. A “no” reached in four weeks with evidence is a good outcome, and far cheaper than a “yes” reached in four months without any.
Mistakes that waste an automation discovery workshop
These five account for most of the disappointing sessions we are asked to review afterwards.
Inviting the whole department
Twenty-five attendees guarantees a presentation. Everyone speaks briefly, nobody is challenged, and the ranking is decided by seniority. Keep it under ten and brief the excluded separately.
Letting the loudest process win
The most irritating process is rarely the most valuable one. Irritation correlates with visibility, not with volume, and an automation discovery workshop that scores on frustration will reliably pick a low-return candidate.
Running it before the data exists
Without volumes there is no ranking, only preference. If the numbers are not available, postpone by two weeks and extract them; a delayed session beats a decorative one every time.
Ending without a decision
The single most common failure. If the day genuinely cannot decide, name the missing evidence, the person fetching it and the date it is due — that is a decision too, and it keeps the momentum alive.
Treating the output as permanent
A shortlist is a snapshot of what you knew on one Tuesday. Systems get replaced and volumes shift, so revisit it every six months rather than defending it for two years.
Measuring whether the automation discovery workshop worked
Treat the session itself as something to be evaluated, using the same discipline you would apply to any other investment.
The ninety-day test
Ninety days later, is something in build, and is it on the shortlist the session produced? If yes, the day worked. If what is being built never appeared on the list, the ranking was ignored and the next session needs different decision rights.
Leading indicators in the first fortnight
Artefacts circulated within three days, corrections received from at least two practitioners, and the hard technical assumption tested. All three are cheap to observe and predict the ninety-day answer well.
Lagging indicators at six months
Did the benefit case survive contact with reality, and is anyone measuring it? Baselines captured before the build are the only way to answer honestly, and the discipline in the GOV.UK Service Manual is a sound model to borrow.
When to run the next one
After the first automation is live and measured, not before. The second automation discovery workshop is far better than the first, because you finally have real build costs and real exception rates to score against instead of estimates.
Frequently asked questions about the automation discovery workshop
How long should an automation discovery workshop last?
One full day for a longlist under twenty-five candidates in a single department. Two days when the processes cross three or more functions, or when several attendees have never mapped a process before.
Do we need an external facilitator?
Not necessarily, but you do need a neutral one. If an internal person can run the day without a stake in which process wins, that works well. Where the candidates cross departmental boundaries, external neutrality is usually cheaper than the alternative.
What if we have no volume data at all?
Postpone. A fortnight of query writing or a light process-mining pass turns an opinion-led session into an evidence-led one, and no facilitation technique can substitute for that.
Should the technology vendor attend?
Not during discovery. Bring vendors in after the shortlist exists, when you can ask each of them to address the same defined problem instead of proposing a problem their product already solves.
How many candidates should come out of it?
Three to five ranked, with one clearly first. A shortlist of one hides the reasoning; a shortlist of twelve has not made a decision.
What if the workshop concludes nothing is worth automating?
That is a legitimate and valuable outcome, usually meaning the processes need simplifying first. Automating an inefficient process only makes you inefficient faster, and finding that out in a day is inexpensive.
References
OMG Business Process Model and Notation 2.0 Specification
GOV.UK Service Manual: Measuring Success
GOV.UK Technology Code of Practice
HM Treasury Green Book: Appraisal and Evaluation in Central Government
AI Playbook for the UK Government
ONS Business Insights and Impact on the UK Economy
ICO Guidance on Data Protection Impact Assessments
ICO Rights Related to Automated Decision Making and Profiling
Microsoft Power Automate: Getting Started
Microsoft Power Automate Desktop Flows
NIST SP 800-30: Guide for Conducting Risk Assessments
