An AI readiness assessment that sits in the work

Companies search AI readiness assessment after a board asked whether they are behind. Vendors answer with quizzes, maturity models, and a red-amber-green slide. Those artifacts can start a conversation. They cannot tell you whether the unofficial spreadsheet is where the hours live, or whether counsel will allow customer text in a model, or whether the first process even repeats. Nidrosoft answers the search with Embed and Diagnose. The leftover is a packet a builder can staff, including a builder who is not us.

We still use the phrase on this URL because that is what people type. The service page for audits is the same work with a clearer noun. If you arrived here, you are in the right place. If you wanted a ten-minute quiz, we will tell you to buy a quiz. We do not sell a fake score.

Cyriac Zeh founded Nidrosoft in San Diego after Anthropic, Microsoft on Bing and Edge, Intuit, and Gap Inc. One hundred twenty-five products shipped. Sixteen public. Readiness in this studio means the company can name a process, an owner, a system, a risk, and a human approval. It does not mean a vibe that leadership is excited.

Systems means the CRM, inbox, practice software, dealer tools, billing, and the personal drive that still closes Friday. Workflows means the repeating path a human walks, including the workaround they will only show you after an hour. Data means what is clean, what is forbidden, and what is duplicated. Risks means customer harm, money, regulated language, and the absence of a signer.

We will ask what people already paste into ChatGPT or Copilot. That paste list is a readiness signal. If the list is emails and notes, you may be ready for seats and training, not a connector. If the list is records they should not have exported, you are ready for a policy conversation before any model work.

We will ask to see model invoices if you already pay providers. TokenTra exists because teams could not see spend by feature. Teams in that class of visibility have cut 10 to 30 percent of wasted tokens. The assessment does not require you to buy TokenTra. It does require honesty about the bill. A company that claims readiness while the invoice is a blob is not ready to add another agent.

Nerlude is a reminder to inventory keys and renewals. Many assessments skip credentials and then a later build dies in security review. We write the key mess into the packet. Protectron is a reminder that EU AI Act language is not a vibe. We separate product artifacts from counsel. Omnira Dental, Eviction Wizard, and Nexuvo remind us that high-care and overnight processes have different readiness bars.

• Named processes with hours
• Systems and unofficial tools
• Data that can move
• Risks that block a write
• Owner and signer
• Recommended next buy

Week one is Embed. We join operators. We watch. We sit with the owner of the system of record. Without that owner, readiness will be fiction. Week two is Diagnose and a first Scope sentence for the top process: what would run, what a person would approve, what you would measure.

You receive a ranked map, risk notes, open questions that would change the rank, and a recommended next purchase. The next purchase might be seats. It might be a leadership sprint. It might be a scoped build. It might be a boutique bake-off. It might be a pause. A pause is a valid assessment outcome.

Remote assessments are normal. San Diego, La Mesa, and Southern California teams can ask for on-site days. Cyriac travels when the floor requires it. We will not invent a local readiness office in your city.

If you only have a one-hour steering slot, you are not buying this assessment. Buy a workshop to create permission, or buy a survey from someone who sells surveys. Squeezing Diagnose into a single call reproduces the loudest opinion in the room.

Surveys win when you need a slide tomorrow and you accept that nobody sat on the floor. Strategy diagnosis wins when the audience demanded a named firm’s pack. Vendor checklists win when the money will go to that vendor anyway. Seat trials win when drafting is the hypothesis. Our assessment wins when someone will later build, train, or deliberately stop with a written reason.

Price stays off the page. A two-week embed is not a quiz. Book the calendar. We will not invent a readiness fee in euros to look complete.

Ready for a build means a repeating process, a named owner, a system we can reach or copy, a signer for consequential actions, and a measure the department already believes. Ready for training means people can bring tasks and a willingness to fail in the room. Ready for seats means an admin, an acceptable-use note, and curiosity. Ready for a strategy pack means a board date and no write path yet.

A company can be ready for seats and not ready for a build. That is the most common honest result. Excitement is not readiness. A chatbot mockup is not readiness. A transformation banner is not readiness.

Under 30 people, readiness includes a founder who will live in the first approval. Past 250, readiness includes a sponsor who will protect one department. Between 30 and 250, readiness includes a lead who can open the tool without a three-week ticket.

NanoBrief shows a document job collapsing from three-plus hours to under two minutes when the questions are known. That product is ready because the brief process repeats. Your assessment will look for the same repetition in your nouns.

You own the ranked table, the system notes, the risk list, the open questions, the drafted prompts from sessions, and the recommended next buy. You do not own a production system unless you buy a build. You do not own a playbook unless you buy training. You can send the packet to a shop, a boutique, or an internal lead.

We write in sentences a new COO can read aloud. We name tools by the names you use. We name people by role. If a page of the packet cannot survive that test, we rewrite it before we leave.

Octana, Guidera, Clear Audio AI, and MixGuidera may appear as examples when the work is media or travel. They are illustrations. Your packet should still speak in claims, chairs, lots, SKUs, or funds.

We will not sell a quiz dressed as an audit. We will not sell a two-week theater if security has already said no environment exists this half. We will not sell an assessment whose only audience is a press blog. We will not sell one that must conclude you are ready because a budget is waiting.

We will not sell one that ignores the unofficial stack. We will not sell one that replaces counsel. We will not sell one for a process that happens once a year. We will not sell one if no operator can attend.

The packet may send you to Copilot or ChatGPT for a quarter. It may send you to corporate training so leaders share a language. It may send you to a Nidrosoft build. It may send you to a product shop because the leftover is a specified app. It may send you to a strategy firm because the leftover is still a pack. Those recommendations are the product.

If the top process resembles public work, we will say so without pretending you will receive that product. Inventory nights resemble Nexuvo (3x inventory, 80 percent less manual outreach). Practice operations resemble Omnira Dental. Spend blobs resemble TokenTra. Notice files resemble Eviction Wizard. Briefs resemble NanoBrief. Resemblance is a hint for Scope, not a SKU.

Transformation is a later word. An assessment that jumps to a company-wide banner has failed Diagnose. Team one first. Always.

We assess with the same eyes we use to build. That is why the method starts with Embed. A researcher who only interviews executives will miss the Friday export. A builder who never writes a map will jump to an agent that misses the hours. Nidrosoft does both, and will stop after the map if you want.

Public systems: Omnira Dental, Octana, Eviction Wizard, Guidera, TokenTra, Protectron, Nerlude, Nexuvo, NanoBrief, Clear Audio AI, MixGuidera. Sixteen public. One hundred twenty-five shipped. Twelve-plus years. Those facts bound what we will claim. They do not invent your score.

The AI Architect is the practice name. Nidrosoft LLC is the company. Cyriac Zeh is the person in the interviews.

Operator, system owner, decision maker. Sometimes one founder. Sometimes three people who dislike each other. We need all of the functions. Legal and security join if a write is on the table. They do not replace the operator.

Contractors who close the unofficial path should be invited by name after day one. Shift workers need a slot that matches the shift. If the only available people are a steering committee, you are buying a strategy conversation. Hire the firm that sells those.

Recordings happen only if you ask and counsel agrees. Notes are usually enough. Operators talk more clearly when they are not performing for a file that might be forwarded to a vendor they distrust.

Write cyriac@nidrosoft.com or use the public calendar. Say that you searched readiness. Name the department and the tool. Attach a survey you already ran if you have one. We will treat it as a hypothesis, not as a fact.

If a strategy firm already delivered a maturity model, send the pages. We will mark which items are processes we can validate on the floor. We will not rewrite their pack for sport.

Cyriac Zeh is in San Diego. The assessment follows the systems. No invented office. No invented score.

Bring the unofficial path on day one if you already know it. The export. The personal inbox. The night someone retypes a listing. Readiness that only interviews directors will grade culture and miss hours. We would rather look rude asking for the spreadsheet than polite while writing a fake green score.

If IT already ran a Copilot pilot, send the useful chats and the failed chats as a pair. That pair is a readiness artifact. A usage graph without examples is not. We will use the pair to decide whether the company is ready for seats-plus-practice or ready for a write path. Those are different next buys.

If procurement asked for a maturity number so they can compare vendors, we will still refuse to invent one. We can give you a ranked table of processes. A vendor who offers a 73 out of 100 is selling comfort. Comfort is not a map. Take our packet to the bake-off and make the other vendors respond to named workflows.

Octana, Guidera, Clear Audio AI, and MixGuidera only change the assessment if your work is media or travel. A manufacturer should not receive a travel-companion lab because those products exist. Your packet stays in your nouns. The public archive bounds what we will claim, not what we will force onto the map.

A company can be ready in one department and unready in another. Write that split into the packet. Enterprise banners that average the split into a single color are how good desks wait a year. Department one can start. Department two can stay on seats. That is readiness with a spine.

If the board asked whether you are behind competitors, we will still refuse a percentile. We can tell you whether a repeating process has an owner and a path into the tools. Behind is a story word. Hours and access are assessment words. Use the story word in a strategy pack if you hire that firm. Use the hours in our packet.

Protectron, TokenTra, and Nerlude only enter the packet when compliance artifacts, token bills, or credentials showed up in the interviews. We will not pad readiness with products you do not need. A clean packet is shorter and more useful than a tour of the archive.

If you want a score for a slide this week, buy a survey from someone who sells surveys. Come back when someone on the floor can open the unofficial stack. Mixing those buys in one week wastes both. A green number that nobody can staff is still a poster.