Contextro turns workflows into structured context for AI implementation.

A framework for translating messy business workflows into a structured specification that AI systems can actually work with.

See how it works →
A human hand and a robot hand exchanging a Contextro workflow blueprint

Most AI implementation starts in the wrong place.

Organisations begin with tools, models, prompts, or agent ideas. They move fast, experiment broadly, and get limited results. The usual explanation is that the AI is not good enough, or the use case is not quite right.

The actual issue is more fundamental. Workflows inside real organisations are undocumented in the way they actually run. They are spread across people, tools, files, and informal decisions. They depend on rules no one has written down and judgment no one has explained. They have no clear definition of what a good output looks like.

AI systems cannot operate effectively in that kind of ambiguity. They need defined inputs, explicit rules, known constraints, clear outputs. Without that structure, even well-chosen AI will underperform.

The gap is not in the AI. It is in the workflow.

Workflow as it typically exists

sarah usually handles the intake emails

if it's urgent someone pings the channel

docs go in the shared drive somewhere

we check with legal if it looks risky

output depends on who's doing it

turnaround is "a few days usually"

Contextro specification
trigger: inbound email to intake@domain.com
escalation_rule: priority = "urgent" → notify #ops-critical
storage: /intake/YYYY-MM/[case-id]/
compliance_check: risk_score > 0.7 → legal_review = required
output_schema: IntakeReport v2.1
sla: 48h from trigger to output

Contextro is a context-driven workflow specification framework.

It is designed to take workflows as they actually exist (informal, fragmented, tacit) and translate them into structured context that AI systems can work within.

The key concept is context. In Contextro, context does not mean background information. It means the full working structure an AI system needs to operate reliably inside a workflow: its goals, inputs, rules, decision logic, system access, memory requirements, human checkpoints, and success criteria.

A process map shows what happens. A Contextro specification defines what an AI system would need in order to participate in, support, or run that workflow safely and effectively.

That distinction — between process description and context specification — is what makes Contextro different.

“A process map shows what happens. A Contextro specification defines what an AI system needs in order to operate.”

The Thirteen Context Dimensions
01GoalsWhat the workflow is trying to achieve and why it matters
02InputsWhat information, data, or materials enter the workflow
03OutputsWhat the workflow produces, in what form, and to what standard
04InstructionsThe procedural steps, sequences, and methods used to do the work
05RulesThe explicit and implicit rules that govern how the work is done
06Decision LogicWhere judgment is required, what factors inform decisions, and what the decision boundaries are
07System AccessWhat tools, platforms, databases, and integrations the workflow touches
08Workflow StateWhere the workflow is at any given point: what has been done, what remains, what is blocked
09MemoryWhat the workflow needs to remember across sessions, handoffs, or iterations
10DependenciesWhat must happen before, after, or alongside this workflow for it to function
11ConstraintsWhat limits, regulations, policies, budgets, or time pressures apply
12Human CheckpointsWhere human oversight, review, approval, or judgment must be preserved
13Success CriteriaHow the workflow's quality, correctness, and effectiveness are measured

The Contextro process operates across three layers.

Most AI implementation efforts fail at one of three points: starting with the wrong workflow, producing a specification that is never built, or building something that degrades without oversight. The three-layer structure addresses each directly. The pre-process ensures effort goes into the right workflow before analysis begins. The four core phases move sequentially — each only possible because the previous one is complete — from raw workflow to an implementation-ready specification. The post-process closes the gap between that specification and a working, maintained system.

Pre-process
Discover

Map the organisation's work landscape and identify which workflows to examine. The Knowledge Work Map reveals where AI implementation will create the most value, before any deep analysis begins. For organisations without a defined starting point.

Process
01Capture

Document how the workflow actually runs today — not how it is supposed to run. Surface the real steps, the workarounds, the tacit rules, and the informal decisions that never appear in official documentation.

02Decompose

Break the workflow into its constituent tasks and map each one against the full thirteen context dimensions. This is where the workflow moves from narrative description to structured, dimension-by-dimension analysis.

03Assess

Evaluate each part of the workflow for AI suitability and determine the right intervention type: whether to augment, transform, productise, or leave the task fully human-led.

04Specify

Produce the Contextro Spec, the structured context specification for the target-state workflow. Define the target state, human checkpoints, role transitions, and implementation requirements. This is the primary deliverable of the core process.

Post-process
Build

Translate the Contextro Spec into a Build Brief with spec-to-build traceability, and validate that what is built matches what the spec defined before it goes live.

Govern

Define how the implemented workflow will be monitored, evaluated, and improved over time. Implementation without governance degrades.

Contextro produces structured, usable artifacts — not reports.

The outputs of Contextro are designed to be used, not filed. Each artifact has a specific audience and purpose.

01Knowledge Work MapA structured grid mapping an organisation's work across functions and task types, used to identify high-leverage workflows and prioritise where implementation effort will create the most value.
02Workflow BlueprintA structured map of how the workflow actually runs, including tasks, handoffs, decision points, and system touchpoints. Built from what the workflow does, not what it is supposed to do.
03Contextro SpecThe primary artifact: a task-level context specification defining, for each task, the full context an AI system needs to participate in, support, or execute that workflow. Readable by humans; usable directly by AI systems.
04AI Suitability AssessmentA scored evaluation of each workflow task across five dimensions (context availability, decision codifiability, error tolerance, frequency, and human value), informing the intervention classification.
05Intervention Decision MapA task-by-task classification of which parts of the workflow should be augmented, transformed, productised, or preserved as human-led, derived from the suitability assessment.
06Target-State Workflow DesignA concrete specification of how the workflow should operate after implementation, including human checkpoints and role transition definitions for every task where the intervention changes a person's role.
07Build BriefA structured translation of the Contextro Spec into builder-ready requirements, with spec-to-build traceability ensuring nothing in the specification is lost or silently dropped during implementation.
08Governance FrameworkThe monitoring, evaluation, and iteration structures that keep the implemented workflow running well, including metrics, feedback loops, and a plan for updating the spec as the workflow evolves.

These artifacts are designed to be used by humans to redesign work, by builders to implement AI systems, and by AI models themselves as structured operating context.

A framework built on one belief.

AI implementation succeeds when workflows are translated into the right context structure. Not when the right tool is chosen. Not when the right model is deployed. When the workflow itself has been properly understood and specified.

Contextro was developed by Umut Ozturk to make that work more structured, more repeatable, and more defensible. It is a proprietary method and will continue to evolve.

Contextro is where AI starts working.

To learn more about the framework, send me an email or schedule a meeting with me.

Contextro v0.4 — developed by Umut Ozturk