Where it runs
In a managed cloud that AI Depo operates. When your data policy calls for it, the models run on hardware your business owns, and we specify, install and operate it.
Agents
Every operating system AI Depo builds comes with agents. They find what slipped, do the work and show you the result. A person taps approve for anything sensitive.
Book an AI AssessmentAutomations with outcomes
Agents work on the company’s own records, so they can spot the job that never got invoiced and then invoice it. We start where the money leaks: the missed call, the estimate nobody chased, the invoice that went out late.
Real companies' names and data anonymized.
The assistant in the corner
Every screen of the operating system has an assistant in the bottom-right corner. It knows the whole company, it can act on what it finds, and every answer ends with an “Open →” button to the record.
Tap to ask
Pricing, procedures and past work answer the question, and the answer stays current as your information changes.
The same assistant gives a foreman and a CEO different reach, as set by your executive team.
The agent in every pocket
Every employee has their own login on the mobile app, and with it their own company agent. They call or text it for reports and for actions. Here is one example day for a foreman.
Luis PeñaForeman, Crew 3 · Keystone Underground
Before Crew 3 leaves the yard, the agent texts Luis the segment, the open locate and the weather. He texts back a question and it answers from the truck’s load list.
SamI read the schedule, the locates and the forecast before he’s out of the yard.
Gloves on, rig running. Luis calls instead of typing. The agent answers from the locate ticket, then tells Dee to start the bore.
SamHe asked with his hands full. The answer came straight off the ticket.
One voice note at lunch. The agent turns it into production lines on the week’s draft with the photos attached, and catches the rock adder he would have missed.
SamHe talked for a few seconds. Billing gets coded lines and $210 in rock.
Rain is coming. The agent lines up a later start and the texts to the crew, then waits. Luis taps approve and it finishes the job.
SamMoving a crew is sensitive, so I asked first. His role lets him approve it for his own crew.
A short wrap-up of what was logged, what changed and what’s waiting tomorrow. All of it is already on the record for billing and the office.
SamNobody re-types his day. It’s where the office will look for it.
LuisIs the locate at 14.5 cleared yet?
Keystone agentCleared at 9:12. Gas is marked on the east side, about four feet off your line. The photo is in the job.
LuisGood. Tell Dee we start the bore.
Keystone agentDee has it.
Did 340 on C today, rock from about 14.4 on, two photos.
Updates Thursday’s schedule and texts Dee, Ray, Tomás and Ana.
Jobs booked, revenue, open estimates and slow weeks ahead come straight from the operating system, so nobody copies numbers between tools. Anyone can call or text for a fresher cut, inside their own access. When the report shows a slow week, the agent goes and fills it.
Monday report · Crestline Heating & Air · mockup
Next week looked thin. I texted 18 members overdue for tune-ups and booked 14 into the open slots.
Scoped access
Your executive team sets what each role can see and do. The agent works inside those limits on every call, text and screen.
“Fieldstone’s margin is down. Fix it.”Fuel on I-80 was the leak. I re-rated 22 loads. $38,400 back.
“What’s our margin on Route 9?”That’s outside your access. Your footage on Route 9 is 2,480 ft this week.
“Change the Dixon rate to $14.50.”Your login can’t change rates. I’ve sent it to the controller to approve.
“What did Dr. Osei produce last month?”That’s outside your access. Yours was on your Monday report, and I can resend it.
Everyone signs in with their own login, so the approved tool is the easy one.
A simple one your team can read and follow.
Customer and patient details stay inside the system and inside each person’s scope.
We watch what the AI systems touch, and we keep watching after launch.
How approvals work
People decide what matters. By the time you’re asked, the work is finished and laid out, so the decision takes one tap. Try it on the card.
It checks the records, prices the options and prepares the invoice, PO or message.
Anything your executive team marked as sensitive waits for a person, on that person’s phone.
The card shows what will happen and what it costs. “Not yet” is always an option.
The work goes out and lands on the record with the approver’s name.
Bayline agent · Needs your OK
Uponor raised to $36.80 a case. Viega quoted less, so I drafted the order with them.
After launch
Your agents need somewhere to run and someone looking after them. Both are on us, and the code can be yours.
In a managed cloud that AI Depo operates. When your data policy calls for it, the models run on hardware your business owns, and we specify, install and operate it.
Updates, monitoring and fixes after launch, with a person to call when something changes. Your team learns it with a plain guide for every role, and in person on request.
We find where your team loses time and money, and show you what your own agents would take on first.