Try the example or start a new assessment. Enter current-state facts to generate your Executive Dashboard and Executive Brief. Work stays in this page’s memory and resets when you reload. Save keeps a temporary copy in this tab; Export Assessment downloads a file. Live app connections are disabled. Use fictional information.
Direct links for reviewers
Start with Guided assessment or How to use. Vendor evidence and pilot results are optional follow-up work. This demo does not test paid access, persistent customer accounts or live integrations.
THE THREE-STEP GUIDED ASSESSMENT
How to use the Studio
New assessment → short guided factual intake → generated analysis → Executive Dashboard and Executive Brief.
Start with what you know today
No packet is required. Click New assessment and give it a name. Complete these three short sections in Guided assessment. You do not need to visit any other tool to generate your Executive Brief.
- The work. Describe the business problem. The Studio checks the description for a starting AI/tool family using Axiom & Arc’s seven-family decision framework; no packet or external reading is required. Confirm the suggested main task, or choose it when the description is unclear. Factual answers refine the recommendation; they override text suggestions. Then describe how decisions are made and the systems used. Select Unknown when you do not know. Follow-up questions appear only when relevant.
- Current effort. Add one overall task or a short task breakdown, current hands-on minutes, and monthly volume. Exception frequency and labor cost are optional. Enter only what happens today; the Studio calculates future human-effort ranges and recommends how each task could change.
- Constraints & readiness. Answer about data sensitivity, approvals, the pilot owner and a safe manual fallback. Add known constraints if useful, then click View Executive Dashboard.
Your Executive Dashboard and calculated fit range
The Executive Dashboard prominently shows the recommended AI/automation type, its calculated fit range, why it fits and the overall recommended action. Click the grade for the weighted breakdown; click any metric card for the supporting calculations. Open Executive Brief provides the complete exportable output without crowding the Executive Dashboard.
Fit uses authored weights: task match 40%, compatible decision logic 25%, family prerequisites 20%, and safe recovery 15%. Unknown criteria create a percentage range, not a zero or an assumed pass. Fit is not approval probability: controls, risk and evidence determine the separate overall recommendation.
Understanding your AI recommendation
After you describe the issue, the emerging analysis suggests a starting approach when a clear task is recognized. Confirm the factual task description. If several tasks are detected or no clear match exists, choose the main task; the Studio does not guess a winner. You never need to select an AI family yourself.
Read the recommended approach, why it fits, the simpler alternative and the prerequisites. Open How we chose this approach for the facts used, Axiom & Arc reasoning and supporting source links. Unknown controls remain unresolved. An agent candidate with unverified controls stays human-led; predictive work without suitable history first needs a data baseline.
Your generated result
Your Executive Dashboard shows the overall decision status, recommended approach, next action, effort range, estimated available hours and their annual value, unresolved conditions and estimate confidence. Click an Executive Dashboard card to inspect its details, or choose Open Executive Brief for the complete exportable output. The Executive Brief brings together the recommended AI or automation approach, future workflow, effort range, available-time impact, economics where inputs are available, readiness, risks, unresolved conditions, pilot guidance and next action. Open Calculation factors beside a task to see what drove its estimate.
You are never asked to assign a use-case score, vendor score, evidence-strength number or future human minutes. Unknown facts remain conditions to resolve; they are not assumed to pass.
The 10–15 minute basic assessment is a design target, not a measured completion guarantee.
Optional follow-up work
Compare tools: use this only when you have candidate products to compare. Enter requirements, factual results, evidence sources and costs. The Studio calculates the scores and ranking. Vendor comparison is not required for the initial Executive Brief.
Pilot results: after testing, record observed results to support an accountable decision. These are measurements, not predictions you must supply during the basic assessment.
Actions & connections: generate actions from unresolved conditions, assign owners and dates, and optionally send reviewed actions to a configured webhook in the private edition.
Save and revisit
Save unfinished work with a name and valid entered values. Missing information appears in the Executive Brief. Reopen the saved assessment, edit its name or facts in Guided assessment, and save again. Calculations update automatically. Capture a decision snapshot to preserve the current Executive Brief; Export Executive Brief downloads it, while Export downloads the assessment data.
Old manual assumptions may appear in the clearly labeled Legacy manual assumptions section. They are preserved for traceability and excluded from the generated planning model.
What an estimate means
Future minutes are calculated planning ranges, not facts about a vendor. Every range uses disclosed assumptions. No paid AI/API call is required for normal assessment, category recommendations, scoring, estimates or the Executive Brief. This version does not perform full-document AI extraction or live vendor research.
Saving and privacy
Unfinished work can be saved; conditions remain visible in the Executive Brief. In the public demo, records exist only for the current session and reset on reload. Export JSON if you need a copy. The private review edition saves records against the signed-in owner on the server. Publishing the app does not publish those saved assessments. The service operator may have administrative access. Customer-account isolation and commercial access still require launch testing.
Connections
The private version supports outbound webhooks to Zapier Catch Hook, Make custom webhooks (EU1/EU2/US1/US2) and Pipedream HTTP triggers. A receiving workflow can create tasks in apps it supports, such as Asana or Trello; these are not built-in native integrations. Configure and test that workflow yourself. Preview and explicitly send actions; the Studio does not synchronize subsequent changes back. Public demo sends are disabled.
Explore AI and automation approaches
Each family includes best-fit work, when to use it, control watch and representative tools from the approved Axiom & Arc packet. You do not need a separate copy.
Conversational AI
Computer Vision
Document AI + governed workflow
No AI — process improvement or retain current process
Rules / RPA
AXIOM & ARC FIELD GUIDE · PAGE 14
Best-fit work
- Repeatable transactions
- Known decision logic
- Structured handoffs
- Standard approvals
Use it when
- Steps are deterministic
- Inputs are standardized
- Exceptions are limited
- Auditability matters
Control watch
Watch brittle exceptions, hidden workarounds, process drift, and automations that fail the moment reality deviates from the script.
Representative tools
- UiPath
- Power Automate
- ServiceNow workflow
- Business rules engines
Document AI
AXIOM & ARC FIELD GUIDE · PAGE 15
Best-fit work
- Forms and applications
- Scanned or digital documents
- Field extraction and classification
- Large document volumes
Use it when
- Key data sits in documents
- Speed matters
- Standard fields repeat
- Manual extraction is slow
Control watch
Watch low-quality source images, inconsistent document formats, missing fields, and cases where extracted text still needs human interpretation.
Representative tools
- ABBYY
- Google Document AI
- Microsoft Azure AI Document Intelligence
- Amazon Textract
Predictive AI / Machine Learning
AXIOM & ARC FIELD GUIDE · PAGE 16
Best-fit work
- Predictive scoring
- Pattern detection
- Classification
- Forecasting
Use it when
- Historical data exists
- Outcomes repeat
- Accuracy can be measured
- Patterns matter more than rules
Control watch
Watch data quality, bias, drift, overfitting, and whether the model can explain why it reached a prediction.
Representative tools
- DataRobot
- Amazon SageMaker
- Azure Machine Learning
- Google Vertex AI
LLMs / Generative AI
AXIOM & ARC FIELD GUIDE · PAGE 17
Best-fit work
- Drafting and synthesis
- Meeting and document analysis
- Requirements refinement
- Scenario generation
Use it when
- Drafting and synthesis
- Meeting and document analysis
- Requirements refinement
- Scenario generation
Control watch
Watch unsupported inference, sensitive data, inconsistent output, and fabricated specifics.
Representative tools
- ChatGPT Business/Enterprise
- Microsoft 365 Copilot
- Enterprise LLM platforms
RAG / knowledge assistant
AXIOM & ARC FIELD GUIDE · PAGE 18
Best-fit work
- Policy and knowledge retrieval
- Grounded answers
- Search across approved sources
- Evidence-backed responses
Use it when
- Approved sources exist
- Users need answers fast
- Grounding matters
- Traceability is required
Control watch
Watch stale source content, poor retrieval, weak citations, and answers that sound grounded but still overreach the evidence.
Representative tools
- Azure OpenAI with enterprise search
- Google Vertex AI Search
- Enterprise RAG stacks
Digital Twin / Simulation — evaluate fit
DIGITAL TWIN EVALUATION · NIST-ALIGNED EXTENSION
Best-fit work
- Physical assets or operational systems
- Scenario testing and performance prediction
- A defined decision that benefits from model feedback
Use it when
- System behavior can be represented
- Operational data can refresh the model
- Predictions can be checked against actual outcomes
Control watch
A static dashboard or scenario is not a synchronized digital twin. Validate data freshness, model fidelity and decision impact.
Representative tools
- Simulation platforms
- Operational data platforms
- Domain-specific digital twin software
AI Agents — bounded and governed
AXIOM & ARC FIELD GUIDE · PAGE 19
Best-fit work
- Multi-step task execution
- Workflow orchestration
- Bounded actions
- Escalation-aware automation
Use it when
- The task spans steps
- Actions are governed
- Escalation paths are clear
- Human checkpoints exist
Control watch
Watch runaway autonomy, action chaining, unclear permissions, and agents that move faster than the governance around them.
Representative tools
- Microsoft Copilot Studio agents
- Enterprise agent frameworks
- Governed workflow agents
Human expert decision — no AI role established
AXIOM & ARC FIELD GUIDE · PAGE 20
Best-fit work
- Expert review
- High-consequence decisions
- Ambiguous cases
- Exception handling
Use it when
- Judgment matters
- Context is nuanced
- Consequences are high
- The tool should support, not replace, the decision
Control watch
Watch automation pressure, false certainty, overloaded reviewers, and any design that hides the fact that a human is still carrying the real decision.
Representative tools
- Human-in-the-loop workflows
- Review queues
- Decision support platforms
Agentic AI: when should you use it?
Consider an agent when the next step must adapt to the case and the system needs to take bounded actions across a workflow. Several systems or several steps alone do not require an agent.
Use a simpler approach when
- The sequence and rules are predictable: use a conventional workflow, API integration or RPA.
- The job is only to draft or summarize: investigate an LLM assistant.
- The job is only to answer from approved sources: investigate RAG.
Before an agent pilot
- Define permitted actions, least-privilege access and limits on cost and retries.
- Require human approval for consequential actions and define escalation ownership.
- Record actions, test stopping conditions and confirm safe fallback or rollback.
- Test representative cases and demonstrate why a simpler workflow is insufficient.
In this Studio, bounded agentic workflow guidance requires confirmed system access, controls, human checkpoints, recovery and a finding that simpler automation is insufficient. Missing prerequisites keep the recommendation human-led; the category itself is not permission to deploy.
Based on the packet’s AI Agents guidance (p. 19) and Axiom & Arc implementation checks. The workflow-versus-agent distinction is supported by Anthropic’s Building effective agents.
Where the AI recommendations come from
The Studio combines your current-state facts with Axiom & Arc’s authored decision rules. The framework is explained below, so no separate packet, AI expertise or external research is required to use the assessment.
| Nature of the work | Family to investigate | Why |
|---|---|---|
| Handle customer or employee conversations | Conversational AI | The task requires interactive dialogue, intent handling and a route to resolve a request. |
| Inspect images or video | Computer Vision | The task depends on recognizing visual patterns, objects or defects in images or video. |
| Process documents and act across systems | Document AI + governed workflow | The process combines document extraction with validated downstream actions; it needs multiple capabilities rather than one generic assistant. |
| Improve or retain a process without AI | No AI — process improvement or retain current process | A non-AI path was selected, or a tested simpler process already meets the requirement. No AI benefit is established. |
| Move records or transactions | Rules / RPA | Repeatable transactions and structured handoffs suit explicit business rules. |
| Extract from documents | Document AI | Information must be extracted, classified or structured from documents. |
| Predict outcomes | Predictive AI / Machine Learning | Historical patterns can support prediction, classification or forecasting of repeatable outcomes. |
| Draft or summarize | LLMs / Generative AI | The work involves interpreting, drafting, summarizing or restructuring language. |
| Answer from company knowledge | RAG / knowledge assistant | Answers need to remain anchored to approved policies, procedures or company knowledge. |
| Model and simulate a physical system | Digital Twin / Simulation — evaluate fit | A physical or operational system may benefit from a virtual model for testing decisions and predicting behavior. |
| Adaptive system actions | AI Agents — bounded and governed | The next action changes by case, so a bounded agent may coordinate actions across steps. |
| Resolve nuanced cases | Human expert decision — no AI role established | Nuance, ethics or accountability makes human judgment part of the product. |
Methodology and implementation
Axiom & Arc AI Decision Discipline provides the seven-family framework (pp. 13–21), autonomy checks (p. 11) and evidence questions (pp. 23–25). The software implements an authored interpretation of that framework. Conservative keyword rules suggest the main task from your issue; your factual selection takes precedence. This is not a trained classifier or unrestricted natural-language analysis.
References
These publisher sources provide technical background for the approach categories and risk considerations.
- Anthropic. (2024, December 19). Building effective agents. https://www.anthropic.com/engineering/building-effective-agents
- Google for Developers. (n.d.). What is machine learning?. https://developers.google.com/machine-learning/intro-to-ml/what-is-ml
- Microsoft. (n.d.). Introduction to desktop flows. https://learn.microsoft.com/en-us/power-automate/desktop-flows/introduction
- Microsoft. (n.d.). What is Azure Document Intelligence in Foundry Tools?. https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/overview
- Microsoft. (n.d.). Retrieval-augmented generation (RAG) in Azure AI Search. https://learn.microsoft.com/en-us/azure/search/retrieval-augmented-generation-overview
- National Institute of Standards and Technology. (n.d.). Digital twins for advanced manufacturing. https://www.nist.gov/programs-projects/digital-twins-advanced-manufacturing
- National Institute of Standards and Technology. (2023). Artificial intelligence risk management framework (AI RMF 1.0). https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-ai-rmf-10
These sources explain capabilities and risk principles; they do not certify this Studio, rank vendors, validate its classification accuracy or establish its time-saving multipliers. Vendor documentation is technical context, not an endorsement of that vendor. Recommendations are calculated locally from the configured rules, not researched live or sent to a paid AI API.
Use the recommendation to choose what to investigate and pilot. Prove suitability with representative cases, required controls, quality measures and total human effort before deployment. The estimate ranges remain separate, disclosed planning assumptions.
Inspect the decision rules and estimate assumptions
Axiom & Arc planning rules 3.6 · 2026-10-05. These are authored planning rules, not validated forecasts. No paid AI calls are made.
Base future/current time multipliers: Eliminate 0–0; Standardize 0.8–1.0; Automate 0.2–0.6; AI assist 0.5–0.9; Human judgment and approval 0.9–1.1; Exception handling 1.0–1.25. Removal requires explicit confirmation. Automation requires fixed rules, repeatability, documented rules and supported access when systems are involved.
Required approvals set a minimum 0.4–0.8 range. Sensitive or unknown data sets a minimum 0.65–1.0; unknown repeatability, rules or approvals sets 0.75–1.1. Exceptions blend the normal multiplier with at least 1.0–1.25. Unknown exception frequency uses a disclosed 0–30% scenario. Explicit approval and exception tasks retain their own ranges. All factors are assumptions to test, not sourced productivity statistics.
Estimated available hours per month = monthly volume × (current minutes − scenario minutes) ÷ 60. Estimated annual value of available time = estimated available hours per month × hourly cost × 12. First-year net value subtracts recorded annual tool and setup costs. No cash savings are assumed.
The primary category framework comes from Axiom & Arc AI Decision Discipline, approved packet, pp. 13–21, with decision checks on pp. 11 and 23–25. The packet does not establish the numeric timing multipliers. Supplementary category rationale is informed by Anthropic’s workflow/agent distinctions, Microsoft’s RAG guidance and Google’s ML guidance. These sources do not establish the timing multipliers.