Methodology
How Delegate works
The 5-dimension scoring framework that powers every audit.
1. The intake
Four minutes of structured input. We ask for your business profile (industry, size, location), your roles and the tasks each role does day to day, and your pain points (where you wish you had more capacity). That’s it. We ask exactly what we need to score the work — no more.
2. The 5 dimensions
Every task gets scored 0-20 on each of five dimensions. The scores sum to 0-100, which maps to a tier. Two dimensions (Risk Profile and Human Value Displacement) are inverted — lower risk and lower irreplaceability mean a higher score, because they make the task easier to automate safely.
Task Structure (0-20)
How rule-based vs judgment-driven is the work? Deterministic, well-defined tasks score high. Tasks that require context, weighing trade-offs, or reading a room score low.
"Confirming next-day appointments" scores 19 — deterministic trigger, templated message. "Hiring decisions" scores 5 — judgment, context, trust.
Tool Maturity (0-20)
Are there production-ready AI tools that handle this today, or is the category still mostly demos? We score against the actual state of vendor offerings, not roadmap promises.
"Bank reconciliation" scores 17 — QBO, Ramp, and Bill.com all ship mature features. "Permit applications" scores 8 — some tools exist but coverage is narrow and inconsistent.
Integration Feasibility (0-20)
How cleanly does the recommended tool slot into your existing stack? Native integrations and zero-engineering setups score high. "Build a custom integration" scores low.
"ServiceTitan AI Receptionist for ServiceTitan shops" scores 17 — same vendor, no integration work. "Custom-built AI dispatcher on a homegrown CRM" scores 6.
Risk Profile (0-20, inverted)
What’s the cost of getting it wrong? Low-risk tasks (easy to catch errors, low blast radius) get high scores. High-stakes tasks (legal, financial, relationship-critical) get low scores.
"Sending review requests" scores 17 — worst case is a slightly off message. "Negotiating with anchor commercial clients" scores 1 — worst case is losing the contract.
Human Value Displacement (0-20, inverted)
Does automating this actually help, or does it remove value the human provides? Tasks where the human is just doing data entry score high. Tasks where the human IS the value score low.
"Mentoring apprentices" scores 3 — the relationship is the point. "Reconciling bank deposits" scores 14 — the human is just data entry, no irreplaceable judgment.
3. The tier mapping
The 0-100 total maps to one of six tiers. The boundaries aren’t arbitrary — they reflect our judgment of what counts as "production-ready today" vs "promising but not there" vs "should not be automated at all."
| Tier | Score | Meaning |
|---|---|---|
| FULL | 85-100 | AI fully handles this today. Implement and let it run with light auditing. |
| MOSTLY | 65-84 | AI handles 70-90%. A human reviews edge cases. |
| PARTIAL | 45-64 | AI handles meaningful chunks. The human stays primary. |
| ASSIST | 25-44 | AI augments, doesn’t replace. Drafts, suggestions, summaries. |
| NOT YET | 10-24 | Tools aren’t ready. Worth revisiting in 6-12 months. |
| DO NOT | 0-9 | Should not be automated. Legal, ethical, or relational reasons. |
4. The “do not automate” requirement
Every audit must flag at least one task as Do Not. Most flag 3-5. This is a structural feature, not a bug. If your audit doesn’t identify anything that should stay human, the analysis is pattern-matching too generously and the recommendations are unsafe.
Common categories: licensed professional judgment (tax filing, legal advice), sensitive personnel matters (hiring, firing), high-trust client relationships (anchor commercial accounts), and safety-critical work (the actual physical trade in a plumbing or electrical shop).
5. What we don’t do
- We don’t take affiliate fees from any tool we recommend.
- We don’t recommend our own products inside an audit. Delegate is the audit; we don’t sell tooling.
- We don’t sell or share your data. The audit content stays with you and us.
- We don’t add you to a marketing list. The only emails you get are about your audit.
6. What you get
- Per-role deep dive with all 5 dimension scores per task
- Specific tools with current pricing, integration notes, and "best for" guidance
- 90-day implementation roadmap, sequenced by impact and difficulty
- Complete do-not-automate list with category and reasoning
- Industry benchmarks and common pitfalls
- 3-5 specific actions to take this week