Automation programmes fail quietly when they begin with whichever pain was mentioned loudest in last month’s meeting. A structured shortlist forces trade-offs into the open: volume versus risk, excitement versus integration reality. The framework below is designed for workshop use with operations, finance and IT present, not as a substitute for discovery, but as a filter before discovery deepens.
Who should be in the room
Include someone who sees raw customer messages, someone who approves spend, someone who maintains the CRM or helpdesk, and someone who can say no when integration is fantasy. Exclude nobody whose workaround would bypass the finished workflow, often reception or coordinators who hold unofficial spreadsheets.
The six-factor scoring sheet
Score each candidate process from 1 (weak fit) to 5 (strong fit) on every factor. Multiply by the weight if you want emphasis; for most teams an unweighted total out of 30 is enough to rank options.
| Factor | Score 1 (poor) | Score 5 (strong) |
|---|---|---|
| Frequency | Rare ad hoc task | Daily or hourly repetition across the team |
| Rule clarity | Outcomes depend on tacit expert judgement every time | Documented rules cover most branches; exceptions are named |
| Data availability | Inputs live in personal inboxes or paper | Structured fields or parseable documents in systems you control |
| Business impact | Delays annoy but rarely affect revenue or compliance | Direct effect on cash collection, SLA, safety or retention |
| Risk if wrong (reverse) | Errors are irreversible or heavily regulated | Mistakes are cheap to detect and correct |
| Integration readiness | No API, no owner, shadow IT only | Supported connection or export path with a named system owner |
Interpret totals pragmatically: high frequency plus low “risk if wrong” scores often beat glamorous projects that touch regulated data on day one.
Worked example: two candidates compared
Imagine a property services firm choosing between (A) routing maintenance requests from email into a job system and (B) automatically issuing contractor purchase orders.
Candidate A: maintenance intake
- Frequency: 5 (dozens weekly)
- Rule clarity: 4 (urgency categories defined; edge cases escalated)
- Data availability: 3 (email bodies vary; photos attach inconsistently)
- Business impact: 4 (tenant satisfaction and SLA exposure)
- Risk if wrong: 4 (mis-routing is fixable before spend)
- Integration readiness: 3 (helpdesk exists; API partially documented)
Total: 23/30: strong first project with human approval on priority flags.
Candidate B: purchase orders
- Frequency: 3
- Rule clarity: 2 (negotiated rates, caps, preferred suppliers)
- Data availability: 4
- Business impact: 5
- Risk if wrong: 1 (financial commitment)
- Integration readiness: 2 (finance system change-averse)
Total: 17/30: defer until intake discipline and approval paths mature. Property-specific patterns appear in AI Workflow Automation for Property Management.
Signals that a process is not ready
Some low scores are fixable with process design; others are stop signs.
- Two departments disagree on the official steps and both are “right” depending on the day.
- Quality depends on a single expert who has never written down exceptions.
- Success would require bypassing audit trails finance already relies on.
- The proposed automation mainly hides a staffing gap leadership will not address.
Fixing the process, or narrowing scope to one department’s agreed path, often precedes software. Common Automation Mistakes covers what happens when teams skip that step.
Quick wins versus strategic bets
Quick wins score high on frequency and low on integration difficulty: acknowledgements, internal notifications, checklist reminders, CRM updates from web forms. Strategic bets, cross-system reporting, agent-assisted drafting, multi-branch approvals, score high on impact but demand governance. A healthy roadmap sequences one quick win to prove delivery, then one strategic slice with explicit metrics.
Small businesses should resist building both at once with limited admin capacity; see AI Automation for Small Businesses.
Connecting scores to service types
High rule clarity plus moderate data quality often suits Business Process Automation. Messy language with stable escalation rules may need Custom AI Agents for first-pass interpretation. Customer-facing intake overlaps Customer Service Automation and Sales & Lead Automation. Document-heavy admin aligns with Document & Administration Automation.
What to document before requesting proposals
- Named process owner and approximate weekly volume.
- Current tools and where data duplicates.
- Known exceptions and who resolves them today.
- Actions that must never run without approval.
- How you will measure success eight weeks after launch.
That pack makes discovery conversations productive rather than circular. For organisational readiness beyond a single process, read How to Prepare Your Business for AI Integration.
Limitations of scoring
Numbers do not capture change fatigue, union agreements, or customer expectations about speaking to a person. Scores should inform conversation, not automate prioritisation without context. Re-score after the first workflow ships; teams often discover adjacent candidates once data is cleaner.
Running a prioritisation workshop
Block ninety minutes with operations, customer-facing staff, finance and someone who understands IT access, not to build software, but to score three to five named processes on the sheet above. Give each participant sticky notes for exceptions they see weekly; exceptions often reveal hidden risk better than abstract debate.
Facilitation tips: score individually before group discussion to avoid loudest-voice bias; capture disagreements as separate rows if two teams truly run different variants; photograph the whiteboard and attach sample anonymised records so implementers see real messiness.
Weighted scoring when one factor dominates
Regulated environments may double-weight “risk if wrong.” Seasonal businesses may double-weight frequency during peak months only. Document weights explicitly so future re-scoring stays comparable. A spreadsheet with tabs per candidate is sufficient; enterprise governance platforms are not required for SMEs.
From shortlist to discovery brief
Convert the winning candidate into a one-page brief: trigger event, happy path, top five exceptions, systems touched, approvals required, baseline metrics already captured. That brief is what a competent partner uses to estimate feasibility, far more useful than a generic “we want AI” email.
After the winner is chosen
Communicate the decision and the deferred candidates transparently. Teams whose projects scored lower need a sentence on why, not silence, which breeds rumours that automation equals headcount cuts. Tie the chosen workflow to metrics everyone can see in stand-ups.
Revisit deferred candidates
Quarterly, re-score deferred items: integration may have improved, or volume may have grown. Some will remain correctly deferred for years; others become obvious second projects once data hygiene catches up.
Recording dissenting scores
When workshop participants disagree on risk scores, note both views in the brief. Implementers then design explicit human checks rather than averaging away legitimate concern.
Disagreement often marks the border between automation-friendly tasks and those that must remain manual until policy clarifies.
A ranked shortlist is a communication tool as much as a technical one, it explains deferrals and protects teams from endless “quick automation” requests that never score well.
Revisit scores after major business changes, new product lines, acquisitions, or regulatory shifts can turn a deferred project into the right next step without reopening political fights from scratch.
To walk through your shortlist with an implementation partner, request a business consultation. Miguelito AI. Part of Bricat Media, based in Paphos, Cyprus, scopes programmes individually after understanding your real queues and systems.