What an AI Assessment Day Looks Like, Hour by Hour
An AI assessment day is a working session inside your company. We sit with the owners, map the work that eats the most hours, write down the rules any AI system would have to respect, and leave with enough evidence to write you a roadmap. Nothing gets connected, bought or installed that day. The roadmap comes about a week and a half later, and you decide what happens next.
That is the short answer, and it is where a good AI consultant starts. The rest of this post walks through the whole thing: the homework before, the agenda we run on the day, the roadmap you receive, and what an engagement looks like after.
Why companies ask for one
Most of the businesses we work with are successful, established and busy. They are not short on AI tools. They are short on someone inside the company who knows which tools to trust with which work.
The national numbers show how wide that gap is. The U.S. Census Bureau reported in May 2026 that overall business AI use "hovered between 17% and 20%," and that "less than 20% of firms with four or fewer employees reported using AI." Put plainly, most U.S. businesses are not yet using AI in any business function, and the largest firms are well ahead of the smallest.
An assessment exists to close that gap with evidence instead of guesses.
Before the day: two questionnaires
The day works because the homework happens first.
The owner questionnaire. Each owner fills in a ten-page form before we arrive. It covers the company in their own words, where they want it to go, how AI is used today, every piece of software the team touches, the everyday computer work, the bottlenecks, the path a client takes through the company, and where the company's knowledge lives. It opens with a plain statement of how their answers will be handled.
The staff pulse. The team gets a shorter set of questions. Staff see the friction owners rarely see: the file nobody can find, the report rebuilt by hand every month, the email thread that decides a project.
We also read the company from the outside before we walk in: the website, the Google listing, and what search engines and AI answers say about the business today. By the morning of the day, we already have a draft list of where AI could help. The day is for testing that list against the people who know.
The day itself: a four-hour working agenda
This is the agenda we run, close to minute by minute.
0:00 to 0:20. Set the frame
We confirm who is in the room and who is on Zoom. We ask for consent before recording anything confidential, and nothing is recorded without everyone's yes. We state the boundary out loud: assessment first, roadmap second, your choice third, and implementation only under a separate agreement. Then each owner names the one outcome that would make the day worth it.
0:20 to 0:50. Confirm the company's facts
We read back what the public record says about the company, item by item: name, services, the area served, the clients you want more of, the claims you can document. Each one gets marked confirmed, corrected, or needs wording. It sounds basic. It is also exactly what AI systems and search engines need to describe you correctly, so it gets done first.
0:50 to 1:30. Goals, economics and who decides
What has to be different in twelve months? Which three numbers would prove it? What do the current bottlenecks cost in hours, delays and rework? And the questions most AI projects skip: who can approve new software, who can stop a system if it behaves badly, and who owns the upkeep once it is running.
1:30 to 2:30. Map the three highest-value workflows
This is the heart of the day. From the staff answers, we bring a shortlist of candidate workflows. Typical candidates are proposal writing, finding old files, monthly reporting, and following up after meetings. The owners choose three.
For each one we document a recent real example, what triggers it, who does it, where the inputs live, every step and handoff, where it waits, what must stay with a person, what a useful first draft would contain, and how long it takes today. That last number becomes the baseline every later result is measured against.
2:30 to 3:05. Systems, data and boundaries
Where do active files, old files, templates and contracts live? Which copy is the source of truth? Which accounts belong to the company and which are personal? What may never leave company storage? Which client contracts restrict AI use? We never ask for passwords. We record the platform, who controls access, and what approval is needed before anything is connected.
3:05 to 3:35. Is the company's knowledge ready?
AI is only as good as what it can read about you. We check whether written sources exist for the company's positioning, voice, ideal clients, services, glossary, best project stories, standards and checklists. For every gap, we name the person who knows it.
3:35 to 4:00. Read-back and close
We read back the three workflows, the non-negotiables and the human approval points. We confirm the roadmap delivery date. And we say one thing clearly: you can act on the roadmap with us, with another installer, or entirely in-house. It is written to work either way.
After the day: the roadmap
A short follow-up call fills any gaps. Then the written roadmap arrives, about a week and a half after the day.
It is built in ten parts. The ones owners read first:
An executive summary. Three findings, the five highest-value opportunities, what is true 180 days from now, and the three decisions the owners need to make.
A current-state snapshot. The tools in use, the recurring tasks and their cost in hours, and where work waits.
An AI opportunity map. Grouped by area of the business. It includes a list titled What stays human, always, and one titled What we do not recommend now.
An AI use and data policy. Where company information goes today, what is off limits, and secure account setup.
Quick wins. Things the team can use in the first weeks.
A phased plan. Days 1 to 30, 31 to 90, and 91 to 180.
The recommended assistants, each one named for the job it does, plus an estimate of monthly software cost.
A measurement plan, including a 15-minute Monday check.
Implementation options, explained below.
Then we walk through it together in a 60-minute session, so nothing in it is a mystery.
The not recommended list matters as much as the opportunities. A common entry is buying a new tool for a job the team does twice a year. Another is any assistant that would send work outside the company without a person approving it. A roadmap that says yes to everything is a sales document.
What an engagement looks like after the roadmap
The roadmap is the same whoever carries it out. What changes is who does the work. We offer three levels.
Do it yourself | Guided | Done for you | |
|---|---|---|---|
Who builds | Your team | Your AI lead, coached by us | Us, alongside your owners |
Timeline | Your pace | 90 days | Six months of weekly work, then monthly check-ins and optimization for months 7 to 12 |
Best when | Someone inside wants to own it and has the time | You want the skill to grow in-house | You want results with the least senior time |
In every level, the software accounts are yours, in your company's name. Your library of company knowledge is yours too, and you can take it to any AI provider.
How every pilot runs
Before any AI assistant goes into daily use, it gets five things:
An owner who does that work today.
A reviewer with the authority to say no.
A baseline, measured the same way it will be measured later.
A test with the pass mark written down first.
A stop rule. If it does not save time after review, it ends.
The rules that do not bend
Nothing leaves the company unless a person sends it.
Company and client material stays in company accounts.
No pilot starts without an owner, a reviewer, a baseline and a stop rule.
No recording without everyone's consent.
The pace to expect: the phased plan puts the first quick wins in the first 30 days. The larger changes build over the following months, as each pilot proves itself.
What "agentic" means here
You will hear the word agentic a lot. In plain terms, an AI agent is an assistant that does a defined job from start to finish, such as drafting a first reply to a routine client inquiry, filling a standard form from information you already hold, or preparing a weekly summary for the owners. An agentic build is a set of those assistants, each with a narrow job, working from your company's own knowledge, with a person approving anything that matters.
The assessment is how you find out which of those jobs are worth building first. We run our own agency on a team built this way, which is how we learned where the rules above come from.
Working with us
We are Lens on Luxury, based in San Rafael. We run assessment days on-site in Marin and across the Bay Area, and we travel for the right engagement. Our clients include luxury resorts, doctors and surgeons, optical practices, real estate and semiconductor companies: established businesses with no one on staff whose job is AI.
The assessment is a fixed fee you know before the day is booked. Implementation is quoted after, once the roadmap shows what is worth building. If you would like to talk it through first, you can book an hour with me.
Tracey Bauer Founder, Lens on Luxury
Frequently asked questions
What does an AI consultant actually do? A good one starts by learning how your company works before recommending any tool. In an assessment that means questionnaires for owners and staff, a working day on-site to map your highest-value workflows and your data boundaries, and a written roadmap with a phased plan, a data policy and a list of what not to do.
How long does an AI assessment take? The on-site working session is four hours. Questionnaires happen in the weeks before. The written roadmap follows in about a week and a half, then a 60-minute walkthrough.
Who needs to be in the room? The owners, or whoever can approve software, spending and process changes. Staff contribute through the staff questionnaire and, where useful, a short group session.
Do you connect to our systems or install anything during the assessment? No. The assessment is for evidence and decisions. Nothing is connected, uploaded, bought or installed until you choose an implementation option under a separate agreement.
Do we have to hire you to implement the roadmap? No. The roadmap is written so your own team or another installer can carry it out.
Do you work outside Marin County? Yes. We are based in San Rafael and work across the San Francisco Bay Area, and we travel for on-site assessments.
