The 5-Point Framework for Deciding Which Workflows to Automate First

Every team has a list of things they wish were automated. The list is usually long, usually vague, and usually ordered by what sounds impressive at a board meeting rather than what would actually deliver measurable value in the next 60 days. The problem is not a lack of candidates. It is a lack of a principled method for choosing between them.

This framework solves that. Score any workflow across five criteria, each on a 1-to-5 scale, for a maximum of 25 points. The numbers will not make the decision for you — but they will force you to be honest about what you actually know about a workflow before you commit to building on top of it.

The goal is not to find the most ambitious automation. It is to find the one most likely to work in production, deliver measurable value, and build team trust in AI-assisted operations.

The Five Criteria

Note that Risk is scored inversely: a score of 5 means failure has minimal consequences, a score of 1 means failure is catastrophic. High risk tolerance = higher score = better candidate. This is intentional — high-stakes workflows are not bad choices forever, but they are bad choices for early systems that have not yet earned trust.

Criterion 1: Frequency

Frequency is the multiplier on everything else. A workflow that runs 50 times a day returns 50 times the value from the same automation investment as one that runs once a week. It also gives you 50 times the data points to measure quality and catch problems early.

Score	Frequency	Example
5	Multiple times per day	Inbound lead routing, support ticket categorization
4	Daily	Daily reporting brief, end-of-day CRM updates
3	Several times per week	Proposal generation, research summaries
2	Weekly	Weekly performance reports, meeting recaps
1	Monthly or less	Quarterly reviews, annual audits

A workflow that scores 1 on frequency is not unautomatable — but it is a bad first choice. The feedback loop is too slow to tune the system, and the time savings are too small to justify early investment. Start with high-frequency workflows and use the learnings to build out from there.

Criterion 2: Pain

Pain has three components: time cost (how many hours per occurrence does the manual process consume?), error cost (how often does the manual process produce mistakes, and what is the downstream impact?), and morale cost (is this the kind of work that drains the people doing it?). A strong pain score means all three are present.

Score	Pain Level	What it looks like
5	Severe	Hours per occurrence, frequent errors with visible downstream cost, actively resented by team
4	High	30–90 min per occurrence, inconsistent quality, a known bottleneck
3	Moderate	15–30 min, manageable quality, mildly tedious but not resented
2	Low	Under 15 min, reliable quality, people do not think much about it
1	Minimal	Quick and easy already; automation adds complexity without meaningful benefit

Be honest about error cost. Teams often underestimate this because errors in internal workflows are invisible — they do not generate tickets or complaints. A qualification agent that miscategorizes 20% of leads does not create a visible fire. It silently degrades revenue. That counts as high pain.

Criterion 3: Structure

Structure is the criterion teams most often overestimate. A workflow feels structured when you describe it in a meeting. It feels much less structured when you try to write it down step by step and realize three people do it differently, two of the inputs exist in inconsistent formats, and nobody can explain what happens in the edge cases.

The test: can you write a complete process document for this workflow in under an hour that a new hire could follow on day one without asking clarifying questions? If not, the workflow is less structured than you think.

Score	Structure Level	What it means
5	Fully documented	Trigger, inputs, steps, decision points, outputs, and exception handling are all written down and consistent
4	Clear in practice	Not written down, but everyone does it the same way and could describe it precisely
3	Mostly consistent	Core process is agreed-upon; some variation in how edge cases are handled
2	Variable	Different people handle it differently; some tribal knowledge involved
1	Undocumented chaos	No consistent process; relies heavily on judgment, context, and institutional knowledge

Workflows scoring 1 or 2 on structure need to be standardized manually first. Automating a chaotic process does not clean it up — it makes it faster at being chaotic. Do not skip this step.

Criterion 4: Risk

Risk is scored inversely because it acts as a ceiling on appropriate automation maturity. A high-risk workflow is not a good candidate for an early, untrusted system — not because it cannot eventually be automated, but because the cost of failure before trust is established is too high.

Ask two questions: how bad is the worst-case failure scenario? And can a human catch it before it causes damage?

Score	Risk Level	Failure looks like
5	Very low risk	Internal draft only; human always reviews before any external action
4	Low risk	Goes external but impact of error is minor and easily corrected
3	Moderate risk	Wrong output wastes time or requires rework but does not damage relationships or revenue
2	High risk	Wrong output could damage client relationships, trigger complaints, or affect revenue directly
1	Critical risk	Wrong output affects safety, compliance, finances, or cannot be undone

Workflows scoring 1 on risk are not unautomatable — they require mature systems with proven track records, robust exception handling, and explicit human-in-the-loop steps. Build trust with lower-stakes workflows first, then graduate to higher-stakes ones once the system has a track record.

Criterion 5: ROI

ROI is the most important criterion to be concrete about. Vague ROI claims — "saves time," "improves efficiency" — are not good enough. A real ROI case answers: how many hours per week does this save at what blended hourly cost? Does it reduce error rates that have a measurable downstream cost? Does it increase throughput in a way that directly ties to revenue?

A simple formula for time-based ROI: (hours saved per occurrence) × (occurrences per week) × (blended hourly rate) × 52 = annual value. A workflow that saves 45 minutes, happens 10 times a week, and involves people earning the equivalent of $50/hour generates $19,500 per year in recovered labor. That is a concrete number worth building for.

Score	ROI Level	What it means
5	Very high	Clear annual value over $20k, or directly enables revenue at measurable scale
4	High	$10k–20k annual value, or significant quality improvement with visible customer impact
3	Moderate	$3k–10k annual value, or frees capacity for higher-leverage work
2	Low	Under $3k annual value; improvement is real but small
1	Unclear	Cannot articulate the value in concrete terms

How to Apply the Scores: Three Worked Examples

Example A: Inbound Lead Qualification

A B2B SaaS company gets 60 inbound leads per week. A sales rep spends 20 minutes manually reviewing each one, deciding whether to route to a senior rep or handle with a templated response. Wrong routing happens about 15% of the time, wasting senior rep time or losing warm leads.

Criterion	Score	Reasoning
Frequency	5	Multiple times daily
Pain	5	20 min × 60/week = 20 hours/week; 15% error rate with direct revenue cost
Structure	4	Clear qualification criteria exist; some edge cases handled variably
Risk	4	Routing errors are recoverable; no direct client communication without human review
ROI	5	20 hrs/week at $40 blended rate = $41,600/year; plus routing accuracy improvement
Total	23 / 25	Build this first.

Example B: Weekly Investor Update Report

A startup founder spends 3 hours each Monday compiling metrics from Stripe, HubSpot, and Google Analytics into a formatted investor update. The format is consistent; the inputs are well-defined.

Criterion	Score	Reasoning
Frequency	2	Once per week
Pain	3	3 hours/week; tedious but low error rate
Structure	5	Same format, same sources, same metrics every time
Risk	3	Goes to investors — quality matters, but errors are catchable before send
ROI	3	3 hrs/week at $100/hr founder rate = $15,600/year in recovered time
Total	16 / 25	Build after the higher-frequency wins.

Example C: Client Contract Drafting

A consulting firm wants to automate first-draft contracts. Contracts vary significantly by client type, engagement scope, and jurisdiction. Errors could create legal exposure.

Criterion	Score	Reasoning
Frequency	2	3–5 per week
Pain	3	2 hours each; tedious but not error-prone in the current manual process
Structure	2	High variability by client and context; significant judgment required
Risk	1	Legal exposure; client-facing; errors may not be caught before damage
ROI	3	Meaningful time savings but unclear accuracy improvement
Total	11 / 25	Do not automate yet. Standardize contract types first, build trust on internal workflows, return when structure score improves.

What to Do When Two Workflows Score Similarly

If two candidates are within 2–3 points of each other, apply these tiebreakers in order:

1. Pick the higher-frequency one. More runs means faster learning, more data for quality measurement, and a faster path to proven ROI.
2. Pick the one with a clearer owner. An automation that nobody owns in production does not get maintained, improved, or adopted.
3. Pick the one whose failure is more recoverable. When in doubt, start with the workflow where getting it wrong is cheaper to fix.

Using the Framework as an Ongoing Tool

This scoring system is not just a one-time selection tool. Re-score candidates quarterly. A workflow that scored 2 on structure because it was undocumented may score 4 after you standardize it. A workflow that scored 3 on frequency may move to 5 after business growth. The framework evolves with your operations.

It is also useful as a communication tool. When a stakeholder pushes to automate a specific workflow, scoring it together forces a concrete conversation about what is actually known versus assumed. Disagreements about scores are more productive than disagreements about gut feelings.

Frequently Asked Questions

What score is a good candidate for a first automation?

We look for a minimum of 18/25, with no score of 1 on any single criterion. A score of 1 anywhere is a flag that should not be overridden by strong scores elsewhere — a workflow with critical risk exposure or no measurable ROI should not be a first build regardless of how well it scores on other dimensions.

Should I score all five criteria equally?

For most teams, yes. The simple sum works well as a starting filter. If you want to weight criteria differently — for example, if your business has very low error tolerance and risk should count double — build that into your scoring before you run candidates through it, not after you do not like the result.

How do I handle workflows I do not have enough information to score?

Score what you know and flag the unknowns explicitly. A score of "unknown" on structure, for example, is a signal to spend a week documenting the workflow before proceeding. Making a build decision with incomplete information on key criteria is exactly how teams end up with expensive rebuilds.

Can I automate a workflow that scores low on structure if I redesign the process first?

Yes — and this is often the right move. Standardize the manual process first, then re-score. You will frequently find that the act of documenting and standardizing a workflow reveals 30–40% of the automation work has already been done. The automation becomes simpler because the process is simpler.