The 4-question framework a business automation consultant should use on every discovery call to ship in 14 days and pay back in 30.
The mistake we see on almost every first call: the prospect wants to automate the most complex workflow first because it 'has the biggest payoff.' Wrong order. A good business automation consultant scopes the opposite — the simplest workflow that meets four criteria, even if it's not the sexiest one.
The four criteria, ranked
- Repetitive: the same shape happens at least 20 times a week.
- Well-defined: the inputs and outputs fit on an index card.
- Low-risk: a wrong answer is recoverable, not catastrophic.
- High-frequency: the team feels the time loss every single day.
Why these four together matter
Each criterion eliminates a different failure mode. Repetitive eliminates one-off projects that don't compound. Well-defined eliminates scope creep. Low-risk eliminates the need for expensive guardrails on v1. High-frequency makes the team an ally — they feel the win immediately, which buys political capital for the harder builds.
Examples that pass all four
- Inbound lead enrichment + routing.
- Appointment booking from web/WhatsApp/SMS.
- Quote generation from a template + line items.
- Internal FAQ bot for sales reps over your knowledge base.
Examples that fail at least one
- Full proposal generation (well-defined fails — every proposal is custom).
- Refund authorization (low-risk fails — wrong call burns goodwill).
- Annual planning automation (repetitive fails — one shot a year).
- Replacing a CSR entirely (well-defined fails — the job is too broad).
Scoring on a call
On every discovery we score 4–8 candidate workflows on those four axes 1–5. The highest scorer becomes v1 — even when the prospect is emotionally attached to a different one. Sometimes especially when. The job of a business automation consultant is to talk you out of the wrong first build.
The first build's job isn't to impress. It's to prove the model — fast, cheap, boring.
Once v1 ships and pays back, the political and budget cover for the bigger build appears on its own. Skip this step and the second build never happens because the first one over-promised.
The scoring rubric, in full
Here's the exact scorecard we use on discovery calls. Five candidate workflows, four axes, scored 1-5 each. The highest total goes first — and we say no to anything under 12 unless the owner has a strong qualitative reason to override.
WORKFLOW SCORECARD (1-5 per axis)
Workflow | Repetitive | Defined | Low-risk | Frequency | TOTAL
--------------------------|------------|---------|----------|-----------|------
Lead intake + routing | 5 | 4 | 4 | 5 | 18
Invoice chasing | 5 | 5 | 5 | 4 | 19
Appt booking (web/SMS) | 5 | 4 | 4 | 5 | 18
Quote generation | 4 | 2 | 3 | 3 | 12
Annual planning automation| 1 | 2 | 4 | 1 | 8
Full proposal generation | 3 | 1 | 2 | 2 | 8
SCORE BAND | RECOMMENDATION
17-20 | ship as v1
13-16 | scope further, likely v2
9-12 | needs decomposition into smaller workflows
0-8 | do not automate yetThe four questions, expanded
1. Is it repetitive?
Count actual occurrences in the last 30 days. Owner estimates are usually 2-3x reality — pull the data from the CRM, ticketing tool, or call log before scoring. Anything under 20 occurrences/month rarely earns its build cost.
2. Is it well-defined?
Can you write the inputs and outputs on an index card? If the owner needs more than five bullets, the workflow needs decomposition first. The single biggest source of failed first builds we audit is scoping a workflow whose 'definition' was actually a wish list.
3. Is it low-risk?
Wrong-answer cost matters more than right-answer benefit on v1. Mis-routed lead = recoverable. Wrong refund = goodwill burn. Pricing quote = potentially binding under consumer law in some regions. Always start with the recoverable ones.
4. Is it high-frequency?
Daily pain creates a daily proof point. Weekly pain creates a weekly proof point. Monthly pain creates almost no political momentum at all. High frequency makes the team an ally — they feel the win immediately, which buys you cover for the next four builds.
Hands-on: the discovery-call template
This is the actual one-page document we send before every discovery call. The prospect fills it before we get on Zoom, and the call becomes 'let's score these' instead of 'tell me about your business.'
Free 30-min audit
Want us to build this for your business?
We build n8n automations and AI agents for SMBs. Book a free 30-minute audit — we'll map your highest-ROI workflow live, no pitch.
Book a free call# Workflow Discovery — [Company]
## Top 5 workflows that are draining time
1. ...
2. ...
3. ...
4. ...
5. ...
For each, fill:
### [Workflow Name]
- Who does this today? __________________
- How often (per week)? __________________
- How long per occurrence? __________________
- Inputs (data sources)? __________________
- Outputs (what's produced)? __________________
- Worst-case wrong answer? __________________
- Tools touched? __________________
- Owner urgency (1-5)? __________________
## Stack
- CRM: ____ Calendar: ____ Comms: ____
- Forms/chat: ____ Invoicing: ____ Storage: ____
## Constraints
- Compliance: __________________
- Data residency: __________________
- Hand-off rules: __________________The 14-day build cadence
- Day 1-2: scoring + scope sign-off (this document, with totals).
- Day 3-5: integrations wired against staging.
- Day 6-8: AI logic + prompt + tool schema.
- Day 9-10: Slack review channel live, internal soft launch.
- Day 11-13: production rollout, real conversations, prompt tuning.
- Day 14: runbook + handover, retainer kick-off if any.
Why most agencies skip this step
The honest answer: scoping conversations are unbillable and uncomfortable. It's faster for an agency to nod, sell the build, and figure scope out under fire. Talking a prospect out of their preferred-but-wrong v1 is the single biggest signal that you're talking to a serious business automation consultant and not a vendor. McKinsey's State of AI 2025 puts 'misaligned scope' at the top of failed-AI-project causes — across all sizes, not just SMB.
Tutorial: discovery-call SQL queries
Before scoring, we pull data. These three SQL queries cover 80% of the discovery context for any SMB on HubSpot, Pipedrive, GoHighLevel, or any CRM with a SQL-shaped export. Run them, paste outputs into the scorecard, and the conversation writes itself.
-- 1. Highest-volume manual activities (last 90 days)
SELECT activity_type, COUNT(*) AS n,
COUNT(DISTINCT owner_email) AS owners
FROM crm_activities
WHERE created_at > now() - interval '90 days'
GROUP BY 1 ORDER BY n DESC LIMIT 20;
-- 2. Average response time by channel
SELECT source_channel,
PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY EXTRACT(EPOCH FROM (first_reply_at - created_at))) AS p50_seconds,
PERCENTILE_CONT(0.95) WITHIN GROUP (ORDER BY EXTRACT(EPOCH FROM (first_reply_at - created_at))) AS p95_seconds
FROM leads
WHERE first_reply_at IS NOT NULL
AND created_at > now() - interval '90 days'
GROUP BY 1;
-- 3. Leads with NO follow-up touch (silent funnel)
SELECT source_channel, COUNT(*) AS dead_leads
FROM leads
WHERE first_reply_at IS NULL
AND created_at > now() - interval '90 days'
GROUP BY 1 ORDER BY dead_leads DESC;The first query surfaces the repetitive workflows. The second exposes channel-specific response failures. The third is the cheap shocker — most SMBs find 15-30% of inbound leads never received any human touch.
Tutorial: scoping a workflow on a whiteboard in 20 minutes
Step 1 (3 min) Write the trigger on the top-left.
Step 2 (3 min) Write the desired outcome on the top-right.
Step 3 (5 min) Draw arrows between them via the systems
the data flows through (CRM, calendar, comms).
Step 4 (3 min) Mark every arrow that needs a decision —
these are your branches.
Step 5 (3 min) Mark every arrow that touches a human —
these are your hand-offs.
Step 6 (3 min) Count: if branches > 4 OR hand-offs > 2,
decompose into smaller workflows.Three workflows we routinely talk prospects out of
'Build an AI that does X end-to-end'
End-to-end is rarely the right scope for v1. Pick the bottleneck inside X, automate that, ship in 14 days, then expand. The end-to-end version is almost always v3 or v4.
'Replace our CSR / sales rep entirely'
Doesn't work, won't work, and the team will sandbag you for trying. Reframe to 'remove the boring 50% of the CSR's job' and the team becomes an ally.
'Custom AI agent for our unique sales process'
Unique sales processes are usually 'standard sales process + 3 quirks.' Build the standard pattern, then layer the quirks. Building bespoke from scratch costs 4-8x more and breaks faster.
After v1 ships: the second-build decision
The second build is decided by data from the first one. Common patterns:
- If v1 was lead intake, v2 is usually follow-up sequences (compounds the lead-intake gain).
- If v1 was invoice chasing, v2 is usually reporting (the CFO has tasted automation, wants more visibility).
- If v1 was an internal copilot, v2 is usually an external chatbot (proof-point unlocks customer-facing budget).
Failure mode: shipping v1 with no measurement plan
Define the metric on the SOW. 'Reduce response time to under 60 seconds.' 'Recover 20 missed leads per month.' 'Save 8 owner hours per week.' If the metric isn't on the SOW, v2 will be a religious argument instead of a data conversation. McKinsey's State of AI consistently ranks 'no clear success metric' as the top cause of stalled AI programs across company sizes.
What's your AI Readiness Score?
10 questions. A personalized score, profile, and a 90-day automation roadmap built for your business. No email required.
Ready to simplify your business with AI automation?
Let's talk about how intelligent automation can save your team time, reduce errors, and help you focus on what actually drives your business forward.
0 Comments
Be the first to comment. Start the conversation below.