Skip to content
Koderead

Methodology

Seven stages, in order.

The Koderead 7D Method

Most AI projects fail somewhere between 'we should use AI' and 'this runs in production'. The gap is rarely the model. It is that nobody wrote down how the business actually works before automating part of it.

The sequence

Seven stages, in order.

Seven stages, in order. Each one produces something you can read, challenge and keep — whether or not you continue to the next.

  1. 01

    Discover

    How does this business actually make money?

    Revenue model, customer segments, the commercial pressure the CEO is under. We start here because an efficiency gain in a workflow that does not matter is a cost, not a saving.

    Output
    Business context brief
    Typical duration
    2–4 days
  2. 02

    Document

    What do people actually do all day?

    We map the real workflow, including every workaround. The handbook process and the observed process are usually different, and the difference is where the opportunity hides.

    Output
    Workflow map with volumes and time-in-stage
    Typical duration
    1–2 weeks
  3. 03

    Diagnose

    Where does the work actually stall?

    Bottlenecks, rework loops, re-keying, waiting-on-approval time. Measured where possible, estimated with stated assumptions where not.

    Output
    Bottleneck analysis
    Typical duration
    3–5 days
  4. 04

    Determine

    Which of these should AI touch at all?

    Every candidate scored on repetition, human time, data availability, feasibility, risk and revenue impact. Some come back as 'fix the process first' or 'this needs a person'. Those answers are part of the deliverable.

    Output
    Scored opportunity register with ROI ranking
    Typical duration
    3–5 days
  5. 05

    Design

    What does the human-plus-AI operating model look like?

    The future workflow drawn explicitly: what the system does unattended, what it proposes, what stays entirely human, and where the escalation boundaries sit. Signed off before anyone writes code.

    Output
    Target operating model and scope of authority
    Typical duration
    1–2 weeks
  6. 06

    Deploy

    Does it survive contact with real work?

    Build, integrate with the systems you already run, and roll out behind a measured pilot. Adoption is treated as a deliverable, not an assumption.

    Output
    Working system, integrated and instrumented
    Typical duration
    4–12 weeks
  7. 07

    Deliver

    What changed, in numbers?

    Before and after on the metrics agreed at stage 03. If a workflow did not improve, that is reported plainly — we would rather lose the follow-on than defend a result we do not believe.

    Output
    Measured impact report and next-opportunity register
    Typical duration
    Ongoing

Principles

What we hold to when it gets inconvenient.

Understand before building

No code is written until the workflow map is agreed. This is the stage most vendors skip and most failures trace back to.

Automate before adding headcount

The cheapest hire is the one the process no longer needs — but only after the process itself is sound.

Measure outcomes, not features

Shipped is not the same as working. The metric is agreed before the build and reported after it.

Keep humans where judgement matters

Scope of authority is written down and signed. A system nobody can describe the limits of is a liability.

Show the working

Every recommendation carries its assumptions. If we estimated something, we say we estimated it and how.

Build systems that evolve

Configuration is handed over. Your ops lead should be able to change a threshold without raising a ticket with us.

Engagements

Four ways in, priced by scope rather than hours.

The ladder is deliberate: a low-risk analysis first, then one workflow done properly, then breadth. You should never have to buy a transformation to find out whether you need one.

Koderead AI Business Assessment™

Fixed fee

7D stages 01–04 · 7–10 days

A CEO who knows AI matters but not where it would apply here.

Deliverables

  • Business AI Map™ across marketing, sales, operations, service, finance and management
  • Every process classified: keep human, AI-assisted, AI-automated or fully automated
  • AI Opportunity Score per process, using the published eight-factor framework
  • Ranked opportunities with estimated hours and implementation window
  • 90-day roadmap with the sequencing rationale written out

This can conclude that your highest-value fix is not AI at all. That has happened and we reported it. You keep the analysis either way.

AI Workflow Sprint

Fixed scope, fixed price

7D stages 05–07, one workflow · 4–6 weeks

A company that already knows which process is bleeding.

Deliverables

  • Target operating model for the chosen workflow
  • Written scope of authority — what runs unattended, what needs approval
  • Built, integrated and instrumented system
  • Measured pilot with a real before and after
  • Handover, configuration access and team training

Deliberately scoped small enough to finish. One workflow done properly beats five started.

AI Operations Transformation

Scoped per programme

Full 7D, multiple workflows · 8–16 weeks

Fragmented systems, several departments, real data volume.

Deliverables

  • Multi-workflow operating model
  • Agent and automation build-out
  • Systems integration across CRM, ERP and finance
  • Governance, audit trail and access model
  • Management reporting layer
  • Team enablement and adoption tracking

We will usually insist on an assessment first. Quoting a transformation without a workflow map is how these programmes go wrong.

AI Growth System

Scoped per programme

Discover → Build → Grow · 12 weeks, then ongoing

Demand generation is the constraint rather than delivery.

Deliverables

  • Lead capture, qualification and CRM intelligence
  • Sales follow-up automation with human handover points
  • Website conversion architecture
  • Search Everywhere Optimisation — traditional, AI and assistant surfaces
  • Attribution that survives scrutiny

Only worth buying if you can service the demand. We would rather fix delivery first than fill a pipeline you cannot serve.

AI Growth & Optimisation Retainer

Monthly

Continuous 7D · Monthly

Systems already live that need to keep earning.

Deliverables

  • Monitoring of agent and automation performance
  • Prompt, rule and threshold tuning as your business changes
  • Each month: three new automations proposed, one workflow optimised, one cost and one revenue opportunity identified
  • Access, permission and governance review
  • Monthly business-impact reporting

Only worth buying once something is live. We will tell you if a support contract would serve you better.

Stage 01 starts with eleven questions.

The free assessment is a compressed version of stages 01 to 04. It will not measure your actual volumes — only a real engagement does that — but it will tell you where to look.

Start the assessment