Compare
Mireto compared with a general AI assistant
An AI assessment is more than the model behind it. These are five questions to ask of any assessment, and how Mireto answers each.
Five questions
| Question | How Mireto answers |
|---|---|
| Where does the evidence come from? | Research starts from the public record, before the client fills anything in. The Snapshot marks cited text with a bracketed number and lists every numbered source in one Sources list near the end. Client material is named in prose and is not numbered. |
| How is the score made? | A model rates each dimension against a written rubric. Fixed weights combine the ratings. The AI Disruption Score is a number out of 100 and says how much AI is likely to reshape an organization's business model. It does not rate how advanced the organization's own AI adoption is. |
| Whose view does it capture? | The Lens asks the leadership team to rate the six readiness areas that the research rates from public sources, so the team's view sits beside the research. The Lens measures alignment by how far apart the leaders' answers are, not by how high they score. A team that all rates an area low is aligned on it. The readout reads Aligned, Mixed or Fragmented, and lists the areas where answers are furthest apart. Partner Solo and Partner Practice include the Leadership Lens from the day they start. The research seat does not, and the Lens has no price of its own. |
| Which rules does it write to? | The assessment is written to the regulators and policy frameworks of the organization's own jurisdiction, Australia or the United States. |
| Who checks it, and whose name is on it? | A release of the AI Strategic Blueprint needs a quality review first. A separate AI review reads its content against a numbered checklist of failure modes, applies corrections and raises flags for the reviewer. A reviewer can override that step. The AI Strategy Canvas, AI Transformation Roadmap, AI Executive Presentation and cover letter are held until the firm releases them, and the reviewer is recorded. The Snapshot, Analysis, Playbook and Portfolio Assessment are not held. On partner plans, the firm is the named issuer. On the research seat, the issuer is Mireto. |
What the platform adds
- Outside and inside, read together
- The Analysis places each threat and opportunity from the Snapshot in one of five positions, set by whether the client's documents and notes name it and by what the leadership team said: Material and team, agreed; Material and team, split; On paper, team silent; The team sees it, nothing on paper; or Off the radar.
- Cited domains checked
- The domain names that a Snapshot narrative cites are checked against the research the scan gathered, and any the research does not hold are logged for review.
- Business case sums done in code
- In the AI Business Case Generator, payback, ROI, IRR and NPV are computed by fixed arithmetic in code, not by an AI model. Each benefit and cost line states its basis.
- Weekly briefs for partner firms
- Each Monday the platform generates weekly briefs for partner firms: Intelligence, Government and Your World.
Your data
- Will your data train an AI model?
- Our AI providers' commercial terms do not allow them to train on your data.
- Where is our data stored and processed?
- Scan and client content is stored in Australia or the US by the firm's market, or by client country for a firm set up for both markets. It is processed in the US. Mireto's own account, firm, billing and audit records are stored in Australia for every customer. Stripe holds payment records in the United States. A client's record, with its name, website and the intelligence text a partner keeps on it, is stored in Australia, including for US clients. Meeting preps, email drafts, engagement outcomes and relationship briefs for US clients are stored in Australia.
- What does the AI monitoring service see?
- Prompt and output text is never sent to Langfuse, the AI monitoring service. It receives the model, the step, the effort, token counts, cost, the user and firm ids, and error messages.
- Are personal identifiers filtered?
- Text extracted from a document uploaded to a client's library or a scan, and workshop conversation text, pass through a filter for personal identifiers before an AI model reads them. It covers Australian business and tax identifiers, US Social Security and employer identification numbers, and card and account numbers. It does not cover every path. A PDF with little or no text layer is read by the model directly, and an initiative list uploaded for a Portfolio Assessment is not filtered.
- What is not in place today?
- Mireto does not yet have multi-factor authentication, enterprise SSO or SAML, SOC 2, ISO 27001, an external penetration test or a standard DPA.