# Ascension Audit: Map the Business Before You Add AI > The Ascension Audit maps how your company actually works, then decides what to keep, consolidate, automate, augment, build, or leave with people. Includes the seven-level Ascension maturity framework. Source: https://andrewheldy.dev/ascension --- Ascension · the flagship engagement ## The Ascension Audit A structured look at how your company actually operates before anything gets built: the workflows, the software, the knowledge, the people, and the AI already in use, official or not. It usually ends in a prioritized roadmap that says, workflow by workflow, what to keep, consolidate, automate, augment, build, or leave with people. [Book a Discovery Call→](https://andrewheldy.dev/start)[Jump to the framework](#framework) The first step is a 30–60 minute discovery call. The audit itself is scoped after it. The Framework Seven levels, 0 to 6. Describes where a workflow is today. The Audit Maps the operation and determines where each workflow should actually go, which is often not the top. Why audits come first ### Technology should follow the workflow. Choosing a tool before understanding the workflow means guessing at the problem. A team adopts a chatbot, an automation platform, or a specialized assistant because it looked good in a demo, then spends months discovering it solves the wrong problem while the real bottleneck goes untouched. An audit starts with how the business actually runs: where work starts, who touches it, where it stalls, and only then asks what technology belongs where. The goal is better allocation of human attention. Automation is only the means. That is also why the audit comes first: it decides what is worth building before anything gets bought. 1. Business→ 2. Ascension Audit→ 3. Operational map→ 4. Prioritized roadmap→ 5. Implementation→ 6. Training→ 7. Measurement What we map ### Eight parts of one operation. Scope decides depth. A focused audit may look closely at two workflows; a broader one covers the whole operation. Either way, the map describes how the business runs; the software is just where that shows up. 1. 01 #### Workflows How work starts, moves between people, stalls, and reaches a decision. 2. 02 #### Software The systems actually in use, where they overlap, and where information is retyped between them. 3. 03 #### Knowledge Where the answers live, who maintains them, and what the team can trust. 4. 04 #### People Roles, handoffs, judgment calls, and the work people should keep. 5. 05 #### AI Official and unofficial use, what it is helping with, and what is being pasted where. 6. 06 #### Automation Steps that already run on their own, what triggers them, and what happens when they fail. 7. 07 #### Risk Permissions, sensitive information, security boundaries, and accountability. 8. 08 #### Economics Hours, software spend, error and rework, and the baseline everything gets measured against. What we decide ### Six possible answers for every workflow. Each workflow and system the audit maps gets one of six outcomes. The goal is the correct operating model, not the most AI. Two of the six add no new technology at all. 01Keep02Consolidate03Automate04Augment05Build06Keep human #### Consolidate Several tools do overlapping work. Fewer systems, one record, less retyping. Illustrative example One customer record instead of three competing spreadsheets. Levels 0 to 6 ### The Ascension Framework The Framework describes where a workflow is today. The Audit determines where it should actually go. Pick a level to see what it is, what changes, what it requires, and what can go wrong. A company can sensibly run one workflow at level 2 and another at level 5, and a higher level is not automatically a better one. 6AI-Native Operations 5Agents 4Automation 3Connected Intelligence 2Specialist 1Assistantmost businesses 0Search Level 0 is the ground floor. Level 6 is work designed around people, software, and AI from the start. Level 1 · scale 0–6 #### Assistant AI helps one person perform individual tasks: drafting, analyzing, preparing, rewriting, researching. ##### What changes Individual output improves, sometimes a lot. But the business’s workflows have not changed, results vary by person, and every gain disappears the moment the person stops driving. ##### In practice - Draft the reply to a difficult customer email - Analyze a spreadsheet and explain what moved this month - Prepare meeting notes and follow-up actions from a transcript ##### AI role Produces first drafts, analysis, and research on request. ##### Human role Chooses every task, supplies the information each time, initiates every interaction, and reviews every output before it goes anywhere. ##### What it requires - Basic prompting habits - A verification habit before outputs leave the building ##### What can go wrong - Unreviewed output going out under the company’s name - Wildly inconsistent results across employees - Nothing is repeatable; the skill lives in individuals ##### Typical technology General AI assistant, used seriously by individuals ##### Moving up Capture what works: turn the good prompts and context into something reusable. That is the specialist level. See level 2: Specialist↑ [Back to all levels](#levels-overview) Self-diagnosis ### Where are you today? Pick the description that sounds most like your company. Most companies I talk to land in the first band. The audit decides where each workflow should go from here. Which description sounds most like your company? Levels 0–1Employees use ChatGPT, Claude, Copilot, or similar tools individually. Usage is inconsistent and task-by-task, and the gains leave when the person logs off.Level 2Teams have repeatable AI workflows: reusable instructions, company context, or specialized assistants that anyone can run the same way.Level 3AI can securely reach actual company systems and information, so nobody has to paste context in by hand.Level 4+Some processes execute automatically. Work happens because a condition occurred, not because a person remembered to start it. Correct level, not maximum level ### Why more AI is not automatically better The goal for each workflow is the level that fits the work. What sets it is the work itself: - Economics - Reliability - Risk - Reversibility - Complexity - Organizational readiness - Human judgment - Data quality 1. 01 #### Workflows have levels. Companies don’t. There is no single company-wide AI level. Your invoice handling can run at level 4 while your sales research sits at level 2, and that can be exactly right. Assess workflow by workflow. 2. 02 #### Level 6 is the most autonomous rung. The right rung for a workflow may be lower. Some workflows should stay at level 2 forever, because the stakes, the data, or the judgment involved make a person the right owner. Stopping lower is a deliberate decision. 3. 03 #### Use deterministic software for deterministic work. If a step has fixed inputs and one right answer, ordinary software does it cheaper and more reliably than AI. AI earns its place where language, interpretation, or unstructured information creates the value. 4. 04 #### Capability up means control up. Each level trades initiation for oversight. The higher the level, the more the human job becomes permissions, review, exceptions, and accountability. A company that scales capability without scaling control is gambling. The audit process ### Seven stages, from diagnosis to measurement. The first three stages are the audit itself. The rest is what happens if you decide to act on it. Not every engagement uses every stage. 1. 01 #### Diagnose Phase: Audit Starts with the discovery call: clarify the operational problem, which workflows deserve a closer look, and how wide the audit should be. 2. 02 #### Audit Phase: Audit Map the work, the tools, the information, the people, the controls, and the baseline. 3. 03 #### Prioritize Phase: Audit Compare the options by value, feasibility, risk, and who will own the change. 4. 04 #### Pilot Phase: Implement Test one bounded change against real work before widening it. 5. 05 #### Implement Phase: Implement Put the approved workflow, integration, or tool into operation. 6. 06 #### Train Phase: Train Teach the people who will use it and the people who will oversee it. 7. 07 #### Measure Phase: Measure Compare results with the baseline, inspect the exceptions, choose the next iteration. What an audit may produce ### Outputs follow the scope. What you receive depends on the scope we agree before the audit starts. Every output exists to support a decision about the business. Depending on scope, an audit may include - Workflow map - Software map - Knowledge map - AI usage map - Opportunity map - Automation candidates - Implementation priorities - Training requirements - Security considerations - Baseline metrics - Pilot recommendation - Phased roadmap The right tool for each step ### Use the least complex system that reliably satisfies the requirement. A single workflow often needs several kinds of tool. The audit routes each step on its own terms: what it costs, how reliable it has to be, and what happens if it goes wrong. 01Deterministic work02Simple language task03Complex reasoning04Sensitive processing05High-risk action Deterministic work → Traditional software #### Traditional software Fixed inputs, one right answer. Rules are cheaper and more reliable than a model. Illustrative example Total an invoice from validated line items. Treat these as starting positions. The final approach depends on the systems you already run, your security needs, and measured results from a pilot. After the audit ### Implementation follows the roadmap. The audit decides what is worth building and in what order. Implementation builds it, usually starting with one bounded pilot against real work. Sometimes the roadmap says keep what you have. That is a valid outcome, and it costs less than a build. #### Training is part of the job. Training is built on the team’s own workflows and roles, so the system gets used the way it was designed. 1. 01[Audit](https://andrewheldy.dev/ascension)Ascension Audit 2. 02[Implement](https://andrewheldy.dev/services#implement) 3. 03[Train](https://andrewheldy.dev/services#train) ### Start with how your business actually works. The first step is a 30–60 minute discovery conversation. From there, we determine whether an Ascension Audit makes sense and scope it around the actual complexity of your operation: the workflows, the systems involved, the stakeholders, and the depth of analysis it needs. [Book a Discovery Call→](https://andrewheldy.dev/start)[See implementation and training](https://andrewheldy.dev/services) Running this inside a larger organization? [Ascension Enterprise](https://andrewheldy.dev/ascension-enterprise) applies the same audit at organizational scale, where the constraint above level 4 is control rather than capability.