Short answer: List every repetitive process, measure its volume, handling time, error rate, exception rate, how rules-based it is and whether the systems it touches can be accessed by software, then score each one with a weighted table. Do the high-score, low-effort processes first as quick wins, schedule the high-score, high-effort ones as strategic projects, and lay both out quarter by quarter over 12 months. Check data protection (UK GDPR) and Making Tax Digital rules before automating anything that touches personal data or VAT records.
Last reviewed: 3 November 2026
Why a scoring method beats a wish list
Most SMEs already have a list of things people would like automated. It is usually ordered by who complained most recently, not by where the time and risk actually sit. A simple scoring method fixes three problems:
- It compares unlike processes fairly. Invoice matching in finance and order entry in sales can be ranked on the same scale.
- It exposes poor candidates early. A process with many exceptions or no system access looks attractive but often fails in practice.
- It gives you a defensible roadmap. When the board or a budget holder asks "why this first?", you can show the numbers.
The method below needs a spreadsheet and a few weeks of observation. No special software is required, although tools exist to speed it up (see the end of this guide).
Step-by-step method
Step 1: Build a process inventory
Run short workshops (30 to 45 minutes) with each department: finance, sales operations, customer service, HR, purchasing, production planning and quality. For each process, record:
- Process name and owner
- Trigger (an email arrives, a form is submitted, month end)
- Inputs (PDF invoices, spreadsheets, web forms) and outputs (an entry in Xero, Sage or Dynamics, a reply email, a report)
- Systems touched (for example Outlook, Excel, SharePoint, your accounting package, your ERP or CRM)
- Whether personal data is involved
Describe each process at the level of "receive supplier invoice, check against purchase order, post to ledger", not "finance work". If the description does not have a clear start and end, split it.
Step 2: Measure six things
Estimates are fine at first, but replace them with counts from system logs, mailboxes or a two-week tally sheet where you can.
| Measure | What to record | Where to get it |
|---|---|---|
| Volume | Cases per month | System reports, mailbox counts, ticket logs |
| Handling time | Minutes of hands-on effort per case (not elapsed time) | Timed sample of 10–20 cases |
| Error rate | Share of cases needing rework or correction | Credit notes, amended orders, complaint logs |
| Exception rate | Share of cases that do not follow the standard path | Ask the people doing the work; sample cases |
| Rules-based | Can the decision be written as clear rules, or does it need judgement? | Walk through 10 real cases with the process owner |
| System access | Can software read and write the systems involved (API, export, connector), or only a person at a screen? | IT or your software suppliers |
Monthly hours = volume × handling time ÷ 60. This single number is the most persuasive figure in most business cases.
Step 3: Score each process
Give each criterion a score from 1 (poor candidate) to 5 (strong candidate), multiply by the weight and add up. The weights below are a reasonable starting point; adjust them to your priorities, but agree them before scoring so that nobody tunes the weights to favour a pet project.
| Criterion | Weight | Score 1 | Score 5 | How to measure |
|---|---|---|---|---|
| Time saved (monthly hours) | 25% | Under 5 hours | Over 80 hours | Volume × handling time |
| Error and rework cost | 15% | Errors rare and cheap | Errors frequent or costly (credit notes, lost customers, audit findings) | Rework logs, complaints |
| Rules-based | 15% | Mostly judgement | Fully rule-driven | Case walk-through |
| Exception rate (inverted) | 10% | Over 40% of cases are exceptions | Under 5% | Case sample |
| System access | 15% | Screen-only, no export | Modern API or standard connector | IT review |
| Input quality | 10% | Handwritten, free-text, many formats | Structured data or consistent documents | Sample of inputs |
| Compliance and risk | 10% | High-risk personal data or regulated decisions with unclear controls | Low risk, or automation improves control | Data protection and finance review |
Maximum score is 5.0. As a rough guide, processes above 3.5 are strong candidates, 2.5 to 3.5 are worth a closer look, and below 2.5 should usually be fixed or simplified before anyone automates them.
Step 4: Estimate effort separately
Keep effort out of the main score and estimate it on its own scale (low, medium, high), based on the number of systems involved, the integration work and the amount of testing needed. This gives you a two-axis view:
- Quick wins: high score, low effort. Start here.
- Strategic projects: high score, high effort. Plan, budget and phase these.
- Fill-ins: low score, low effort. Do them when there is spare capacity.
- Avoid for now: low score, high effort. Revisit after process changes.
Step 5: Turn the ranking into a 12-month roadmap
Allocate work by quarter, keeping the number of live projects within what your team can support:
- Quarter 1: two or three quick wins, plus discovery and design for the top strategic project. Set up basic governance: who approves automations, where they are documented, who monitors them.
- Quarter 2: build and pilot the first strategic project; deliver further quick wins; measure actual hours saved against estimates.
- Quarter 3: roll out the strategic project fully; start the second strategic project; retire manual workarounds.
- Quarter 4: review results, rescore the inventory (volumes and processes will have changed) and draft the next 12 months.
Worked example (hypothetical numbers)
The figures below are invented for illustration. They are not from a real company or a TechLinkAI client.
A 60-person distributor scores three processes:
| Process | Volume / month | Minutes per case | Monthly hours | Exceptions | Weighted score | Effort |
|---|---|---|---|---|---|---|
| Supplier invoice entry and PO matching | 900 | 4 | 60 | 10% | 4.2 | Medium |
| Sales order entry from customer emails | 1,200 | 3 | 60 | 25% | 3.6 | Medium |
| Monthly management pack | 1 | 960 | 16 | Low | 3.0 | Low |
How the invoice process reached its score: time saved 4 (×0.25 = 1.00), errors 4 (×0.15 = 0.60), rules-based 5 (×0.15 = 0.75), exceptions 4 (×0.10 = 0.40), system access 4 (×0.15 = 0.60), input quality 3 (×0.10 = 0.30), compliance 5 (×0.10 = 0.50). Total 4.15, shown as 4.2 after rounding to one decimal place.
Resulting roadmap for this example: the management pack is a quick win in Quarter 1 (low effort, steady value); invoice matching is the first strategic project (design in Quarter 1, pilot in Quarter 2); order entry follows in Quarter 3 once a standard order template has reduced the exception rate.
UK-specific considerations
UK GDPR and DPIAs
If an automation processes personal data (staff, customers, suppliers who are sole traders), data protection rules apply. Under Article 35 of the UK GDPR, a data protection impact assessment (DPIA) is required before processing that is "likely to result in a high risk to the rights and freedoms of natural persons", and specifically for systematic and extensive evaluation of people based on automated processing, including profiling, where decisions have legal or similarly significant effects [1]. Even where a DPIA is not mandatory, a short screening record shows you considered the risk.
The Data (Use and Access) Act 2025 created what the government describes as a "more permissive framework" for solely automated decisions with legal or similarly significant effects, but with safeguards: telling people about significant decisions, letting them make representations, and offering human intervention [2]. Its measures are being brought into force in stages, so check current ICO guidance before automating decisions about individuals, such as recruitment screening or credit decisions.
Making Tax Digital
Finance automations need to respect Making Tax Digital (MTD). HMRC's VAT Notice 700/22 states that MTD "requires all VAT-registered businesses to keep records digitally and file their VAT Returns using software" [3]. Where data moves between programs that form your electronic VAT records, the transfer must be a "digital link" (for example an API, CSV import or linked cells). HMRC says it does not consider copy and paste to be a digital link [3]. Automations that replace manual re-keying between systems can help here; automations that rely on screen copy-and-paste may not meet the rules.
MTD for Income Tax applies from 6 April 2026 to sole traders and landlords with qualifying income over £50,000 [4]. This matters if you run bookkeeping for owners or associated businesses.
Microsoft 365 environments
If your organisation runs on Microsoft 365, much of the process inventory will involve Outlook mailboxes, Excel files and SharePoint folders. Score system access honestly: a shared spreadsheet that people edit by hand in different layouts is a weaker candidate than one with a fixed structure. Involve whoever administers your tenant early, because permissions and connector policies often decide what is feasible.
Common mistakes
- Automating a broken process. If the exception rate is high because the process is unclear, fix the process first.
- Measuring elapsed time instead of handling time. A request that waits three days but takes four minutes of work saves four minutes.
- Ignoring exceptions. Plan who handles the cases the automation cannot, and how they are flagged.
- Choosing the tool before the process. Score first, then pick technology that suits the top candidates.
- No owner after go-live. Every automation needs a named business owner and a check that it is still working.
- Skipping data protection and finance review until the end, when changes are expensive.
- Never rescoring. Volumes change; revisit the inventory at least annually.
Checklist before you start
- Inventory covers every department, not only the loudest
- Volumes and handling times measured, not guessed, for the top 10 candidates
- Weights agreed before scoring
- Personal data identified and DPIA screening done where needed
- Finance automations checked against MTD digital-link rules
- Effort estimated separately from score
- Each roadmap item has an owner and a success measure
Tools that can help
A spreadsheet is enough for a first pass. Dedicated tools reduce the effort of collecting processes across departments and keeping the scoring consistent. TechLinkAI's Automation Hub finds, scores and prioritises automation opportunities across departments and turns them into a 12-month roadmap. According to the product page, it scores each opportunity by capacity gain, effort and risk and shows them in a catalogue, an impact-versus-complexity matrix and a phased roadmap [5].
FAQ
Which processes should we automate first? Those with a high weighted score and low implementation effort: high volume, clear rules, few exceptions, structured inputs and systems that software can access. Supplier invoice processing and order entry are frequent candidates, but your own numbers should decide.
How long does an automation assessment take for an SME? Inventory workshops and measurement typically take a few weeks, depending on the number of departments and how much data is already available. The scoring itself is quick once the measurements exist.
Do we need a DPIA to automate a process? Only if the processing is likely to be high risk to individuals, or falls within the cases listed in Article 35 UK GDPR [1]. Many back-office automations do not, but record your screening decision.
Can we use copy-and-paste bots for VAT records? HMRC does not treat copy and paste as a digital link under MTD for VAT [3]. Use APIs, imports or other electronic transfers instead.
What is a good score threshold? In the method above, 3.5 out of 5 is a sensible starting line for strong candidates, but the threshold matters less than the ranking and the effort estimate.
Should IT or the business own the roadmap? The business should own priorities and benefits; IT should own feasibility, security and support. Both should sign off the scoring weights.