Audit and strategy that costs the work, then ranks it

Nidrosoft sells audit and strategy as the first paid step. Embed is the first word of the method because the hours only become true when someone watches the workaround. Diagnose is the second word because a list of ideas without payback is a wish list. You can stop after those two weeks. Many buyers should. The map is yours either way. If the leftover is a new app, the next door is Product. If the leftover is a machine in a CRM you already pay for, the next door is AI systems.

The audit looks at systems, workflows, data, and risks. Systems means the CRM, inbox, practice software, dealer tools, billing, and the spreadsheet that still runs Friday. Workflows means the repeating path a human currently walks. Data means what is clean enough to read, what is forbidden to send to a model, and what lives in a personal drive. Risks means customer harm, money movement, regulated text, and the absence of a human on the write.

Cyriac Zeh founded Nidrosoft after design and engineering roles at Anthropic, Microsoft on Bing and Edge, Intuit, and Gap Inc. The studio has shipped 125 products and keeps 16 systems public. We use that archive as a reference for what a later build can look like, not as a promise that your audit will recommend those products. Omnira Dental, Eviction Wizard, and Nexuvo are examples of processes that were concrete enough to ship. Your map may point at something quieter: a follow-up queue, a token bill, a renewal calendar.

People search product consulting San Diego and AI strategy and expect a separate SKU. At Nidrosoft those hours live here. We rank the next spend: a new surface, a machine in a tool you already pay for, seats, a workshop, or a stop. We do not sell a 40-page vision with no hours attached.

If the first leftover is an app, we write the first user and the first job, then point you at Product and the Cost shapes (MVP Sprint, Full Product). If the first leftover is a queue in HubSpot, a practice manager, or a shared inbox, we point you at AI systems. If the company cannot yet share a language, we point you at Workshops.

A board that required a named strategy firm still might need that firm. Our leftover is a process map you can staff. Hire the pack when the audience is the pack. Hire us when someone will later build or deliberately stop.

Week one is mostly Embed. We join the people who do the job. We watch intake, outreach, scheduling, reconciliation, notice prep, inventory hunt, or whatever repeats. We ask what they already tried with ChatGPT or Copilot. The paste history is data. We also sit with the owner of the system of record, even if that person is busy. Without that person, the map will lie about access.

Week two is mostly Diagnose and the first pass of Scope. Each recurring task gets hours, feasibility, and payback. Feasibility includes API reality, data quality, and whether a human approval step will satisfy counsel. Payback includes time returned and error avoided, not a story about innovation. We write the first build as an outcome: what runs itself, what a person approves, what you will measure after thirty days.

You will see us in your tools. Slack, email, the CRM, a screen share on the floor. We do not need a badge in San Diego if the systems are reachable. If you want Cyriac in the room in San Diego, La Mesa, or elsewhere in Southern California, say so when you book. Remote audits are normal for teams outside California.

At the end you receive the ranked map, the risk notes, and a recommended next purchase. That next purchase might be seats and training. It might be a scoped build. It might be a longer transformation program. It might be nothing from Nidrosoft. The last option is a valid audit.

We will also write the questions that would change the rank. A missing API. A works-council date. A vendor who has not answered. Those open questions keep the packet honest when someone reads it in six weeks. If a later builder pretends the questions vanished, you still have the list.

Expect friction in week one. Operators protect workarounds because the workaround is how Friday still closes. We treat that protection as information. The unofficial path is usually where the hours are. A polite tour of the official SOP is useful and incomplete. If we only see the SOP, we will say the audit is unfinished rather than invent hours.

• Named processes with hours attached
• Systems each process touches
• Data that can and cannot move
• Risks that block a write path
• A first outcome written in plain sentences
• Who owns the process on your side

Official stacks are easy to list. Salesforce, HubSpot, a practice manager, a dealer DMS, NetSuite, a shared inbox. Unofficial stacks are the exported CSV, the WhatsApp group, the personal Gmail rules, the night someone retypes Craigslist into the CRM. Nexuvo exists because that night was the job. Your audit will look for the same pattern even if you sell software instead of cars.

We will ask to see Token-like bills if you already pay model providers. TokenTra is our public product for spend by feature, and teams using that class of visibility have cut 10 to 30 percent of wasted AI spend. In an audit we do not require you to buy TokenTra. We do require a look at invoices if “we already use AI” is part of the brief. Otherwise we will diagnose a second copy of a tool you already waste.

Healthcare and dental work means we treat schedule, chart, and patient communication as high-care surfaces. Omnira Dental is live. That does not make every clinic a fit. If you cannot grant a safe environment, the audit will recommend training and seats until you can. Real estate operations such as the notice-to-writ path in Eviction Wizard need the same honesty about legal text and filing authority.

Nerlude is a reminder to inventory credentials and renewals. Audits often find keys in chat. We will write that down as a risk. We will not store your production secrets in a personal vault. If that sentence is a problem for another vendor, keep us.

Every company arrives with a pet use case. A chatbot on the homepage. A voice agent. A magic CRM. Those can win the rank if the hours and the access support them. They often lose to a boring queue: incomplete intake, missed follow-up, inventory that is typed twice, briefs that take three hours. NanoBrief collapses a brief to under two minutes in the public product. Your version might be a scoped generator on your intake fields. The audit decides.

We score feasibility without romance. If the system of record has no API and no export, the idea drops. If counsel will not allow customer text in a third-party model, the idea drops or moves to a private deployment conversation. If the process happens twice a year, the idea drops. Repetition is the engine of payback.

Protectron shows up in audits when someone asks for EU AI Act readiness as a slogan. We will separate product work (classification, documents, trails) from legal opinions we do not sell. The map will say which systems look like they need a trail. It will not pretend Nidrosoft is your counsel.

Octana, Guidera, Clear Audio AI, and MixGuidera appear when the work is media or travel rather than a back office. A marketing team that already briefs video in Slack may need a production path more than a policy. We will still rank it against the inbox that eats Mondays. Boring often wins. That is the point of numbers.

You own the ranked map, the interview notes that are safe to keep, the risk list, and the recommended first outcome. You own the prompts we drafted during sessions. You do not own a production integration unless you buy a build. You do not own a playbook unless you also buy training. Keeping those lines clean is how finance understands the invoice.

If you continue into a build, the audit becomes the first chapter of Scope. We will not re-interview the same people for sport. We will update hours if the floor changed. If you hire a product shop after our map, give them the packet. A shop with a true spec is often the right next partner. We would rather lose the build than watch a shop guess.

The packet is written so a search assistant or a new COO can quote the top three processes. We use plain sentences. We name the system. We name the owner. We name the human approval. If a page of the packet cannot be read aloud in a stand-up, we rewrite it.

• Ranked process table
• System and data notes
• Risk and approval notes
• Recommended next buy (seats, training, build, stop)
• Open questions that would change the rank

Skip the audit if you only wanted Copilot licenses. Buy the licenses. Skip it if the board demanded a McKinsey-style transformation pack and will not accept a process map from a San Diego studio. Hire the firm that lives in that format. Skip it if no owner will attend. Skip it if security has already said no environment will exist this half.

Skip it if you want a tool tour. We will not spend two weeks clicking through a directory of model vendors. Skip it if you want us to bless a build you already promised a board without looking at the floor. We will look anyway, and the map may contradict the promise. That conversation is uncomfortable. It is the job.

Skip it if the company has no recurring work we are allowed to see. A stealth product with no operators yet needs a product shop or an internal founder sprint. Nidrosoft audits operators. Twelve-plus years of shipping does not turn a blank notebook into a process map.

Builds take the first outcome and ship it. Training takes the leadership or engineering group and practices on the tools the map named. Transformation funds the sequence: more builds, more rooms, next team. Renting an AI engineer is a monthly embed when you want a senior seat after the map is clear. Each of those is a different page and a different invoice.

Some maps say the company should stay on ChatGPT or Copilot for a quarter and measure useful versus failed prompts. That is a cheaper education than a connector. We will write it. Some maps say a boutique should run a vendor bake-off because the real need is a platform choice. We will write that too.

If the map points at a process we have shipped in public, we will say so. Inventory acquisition resembles Nexuvo. Practice operations resemble Omnira Dental. Spend visibility resembles TokenTra. Document speed resembles NanoBrief. Resemblance is not a copy-paste. Your approvals and your systems still decide Scope.

Readiness surveys win when you need a number this week and you accept that the number was produced without sitting on the floor. Strategy diagnosis wins when the audience is a board that asked for a named firm. Boutique bake-offs win when the money will go to a platform license. Our audit wins when someone will later build or explicitly decide not to.

The explore page on AI readiness assessment exists for people who searched that phrase. The work they want is often this audit. If they want a score without interviews, they should buy a survey from someone who sells surveys.

Omnira Dental shows a vertical operations surface. Octana shows a media path from a single message. Eviction Wizard shows a legal-operations path with a dashboard. Guidera shows a consumer companion on iOS. TokenTra shows spend control. Protectron shows compliance artifacts. Nerlude shows infrastructure hygiene. Nexuvo shows overnight acquisition with a 3x inventory result and 80 percent less manual outreach. NanoBrief shows document speed. Clear Audio AI and MixGuidera show audio work.

An audit for a fund, a factory, or a retailer will not pretend those products are the deliverable. They are how we prove the studio ships. Your map will name your tools. If we cannot connect a recommendation to a system you showed us, we delete the recommendation.

Formerly Anthropic, Microsoft, Intuit, and Gap is background for judgment, not a claim that those companies are clients of the audit service. Nidrosoft LLC is the contracting entity. The AI Architect is the practice name Cyriac ships under.

We need the operator, the system owner, and a decision maker. Sometimes that is three people. Sometimes it is one founder. Under 30 employees, the founder should be in the room for the ranking conversation. Between 30 and 250, a department lead who can grant access is enough if a sponsor will fund the next step. Past 250, an internal sponsor must protect the department from a company-wide scope creep.

We will talk to legal and security if writes are on the table. We will not let them replace the operator interview. Policies without hours produce bans or blank checks. Hours without policies produce a build that dies in review.

If the only available time is a one-hour steering call, buy a workshop instead. Training can create permission for a later audit. An audit squeezed into a single call will reproduce whatever the loudest person already believed.

Translators, shift workers, and contractors should be planned for. The unofficial path often lives with a contractor who is not on the original invite. If that person cannot join, the hours will be wrong. We will ask for them by name after the first day, once we see who actually closes the work.

A note on recordings: we only record if you ask and counsel agrees. Notes are usually enough. Operators speak more clearly when they are not performing for a file that might be forwarded to a vendor they dislike. The packet should still be readable without the audio.

Use the public calendar or cyriac@nidrosoft.com. Say which department, which tool, and which weekly pain. If you already have a strategy pack, attach it and tell us which page you were told to implement. We will mark what an audit can validate and what still belongs to the firm that wrote the pack.

Scope and price come after we understand access and volume. That is why this page has no retainer and no invented currency figures. A 40-person clinic and a 200-person software company are not the same two weeks.

Cyriac Zeh is in San Diego. We will travel or stay on the link as the work requires. We will not invent a local office in your city to win the booking.