Andrew Duggan

Whose account is your lab's work in?

Watch on YouTube

Picture a quiet Tuesday evening at a dental lab I have invented, so nobody gets blamed. A technician cannot read a dentist’s prescription, so she pastes it into her own AI account with the case notes. A tidy summary comes back. When a dentist’s office asks for a quote, she pastes in the lab’s price list as well. She copies the answers across and logs off.

Next morning the owner opens the books and nothing looks wrong, because on paper nothing is wrong. What the books cannot show is where the prescription, the case notes and the price list went: into a personal account the lab does not own and cannot open. That is the blind spot.

How common is it?

A large survey found that 78% of knowledge workers who use AI at work bring their own AI tools. At small and medium-sized companies, it is 80%. In a separate survey of privacy and security professionals, 48% admitted entering non-public company information into AI tools. These are people whose job is guarding data.

What it costs, in this example

Every figure here is my assumption, for the invented lab. Say 6 technicians use their own AI accounts for lab work, and each loses 1.5 hours a week finding or redoing work held there. Say a technician’s hour costs $40 with wages and overheads. Say 3 technicians leave in a year, each leaving 20 hours of notes and templates to rebuild. That comes to $1,760 a month, or $21,120 a year: worked out, not a result anyone got. It counts only time, and puts no price on the lab’s data sitting under another company’s terms.

The usual fixes

Hiring someone to set rules helps, but nobody hired can see what goes into an app on a technician’s own phone. A written rule helps too; in that privacy survey, 63% said their organisation had set limits on what data can be entered. But a busy bench reaches for the quickest tool. An app for the whole lab is better than nothing, yet without steady checking the work can still sit under the app company’s terms.

What I build instead

This is my method, not a client result. The agent runs inside the owner’s own cloud account, so the prescriptions, case notes and price list stay where the lab can open them. It does only the tasks the owner has approved. When it drafts a quote reply, the reply waits: nothing it writes reaches a dentist or patient until the owner, or a person the owner names, approves it.

Say a message asks for a patient’s full case history at a new address. The agent drafts the send, then stops, because the address is not on file. The owner sees that on the card, taps hold, and rings the dentist’s office on the number the lab already has. The request, the flag and the hold are each written down with the time, in the lab’s own account.

You stay in charge of what the agent may do, what goes out, and the record of it all.

If you would like to see where your staff’s work and your customers’ data really end up, book a blind spot review: 30 minutes, $250, with me. I will tell you plainly whether an agent would help.

Sources

The business shown in this video is an illustration.

Narration: an AI voice. Researched and drafted with AI; reviewed and approved by Andrew before publishing.

Shorts from this video

  • The work that left the lab, in a minute

  • What personal AI accounts cost in time

  • The case history the owner held back

The blind spot review

30 minutes with me, on your business. You tell me how work comes in and where it stalls. I tell you where you’re losing the most and what I’d fix first.

Afterwards you get one page: your three biggest blind spots, in order, with what each fix involves. It is yours whether or not you hire me.

$250

30 minutes · credited in full if you hire me within 30 days

Pick a time