AI audit and diagnosis
We measure your repetitive processes, find where AI brings the highest ROI and deliver a report with estimated savings and an implementation plan.
Typical result
Report in 2-4 weeks, with savings quantified per workflow.
AI Business Tuning
We measure what repetitive work costs you.Then we automate it.
AI agents take over up to 90% of your repetitive workflows, and you pay us only a share of the savings we prove.
30 minutes, free. We talk about your processes, not about slides.
We plug into what you already use
- Excel
- e-Invoicing
- ERP
- Accounting
Before / manual
After / automated
01 / Problems
Repetitive work isn't on your balance sheet.But you pay for it every month.
The same tasks, every day
Your people spend hours copying data from one place to another, filling in the same documents, checking the same things. That's not why you hired them.
Errors cost more than you think
A wrong invoice, a duplicate order, a missed deadline. Every manual error costs time, money and your clients’ trust.
You're growing, but the team can't keep up
More clients, same people. Instead of hiring, you can multiply the capacity you already have.
You don't know where to start with AI
Everyone talks about artificial intelligence, but nobody explains what it can actually do for your company, with your processes and your people.
02 / Services
Four services, one goal.Hours recovered, measured and proven.
- Details
AI agents for processes
We build AI agents that take over 90% of repetitive tasks: documents, invoicing, quotes, reporting, internal communication.
DetailsTypical result
40-80% less time on automated processes, zero copy errors.
Data analysis and dashboards
We turn the data you already have into decisions: visualisations, automatic alerts, reports that build themselves.
DetailsTypical result
Monthly reports: from 2 days of work to 10 minutes of automatic generation.
AI training for teams
Hands-on workshops, not theory. Your people leave knowing how to use the tools, not just what they are.
DetailsTypical result
Team working independently in 1-5 days, with reference material and 30 days of support.
03 / How we work
From stopwatch to agent in production.Four steps, no operational disruption.
01
Diagnosis
We come, observe and measure. We time the workflows, talk to people and find where hours and money are lost.
Duration:2-4 weeks
02
Prototype
We build a working version on the highest-impact workflow. We test it with your team and adjust.
Duration:2-4 weeks
03
Implementation
Old and new run in parallel, gradual migration, team training. Zero operational disruption.
Duration:2-6 weeks
04
Support and tuning
Monitoring, adjustments, monthly impact reports. You pay from the savings we prove.
Duration:Ongoing, 48 months
04 / Pricing
We don't sell consulting hours.We sell results, and you pay from them.
We measure the workflow before, calculate the savings after and take a share of the difference. If you don't save, you don't pay.
Performance contract
5-20%
of proven savings
48-month contract
- We measure the baseline: how many hours and how much money the workflow costs today
- Savings are calculated monthly, on a dashboard you can see too
- The percentage depends on workflow complexity and is agreed before we start
Alternative
Pay upfront, 25% off
Prefer a fixed cost? Pay the contract value upfront, 25% lower.
AI cost
Tokens at cost, zero margin
AI model usage is billed exactly as consumed, with no markup. You see monthly usage on the dashboard.
Worked example
- Measured savings / month
- €600
- AIBT Labs fee (15%)
- €90
- You keep / month
- €510
Illustrative figures. The actual percentage is set after the audit.
Illustrative example
Expense claims from field staff
Installation company, 15 field technicians with cash advances, ~300 receipts a month
Technicians buy small things on the way: fittings, sealant, a replacement drill when one breaks on site, parking. At the end of the month they hand in a pile of crumpled receipts, some faded from the heat. The bookkeeper types them into Excel one by one, matches each to a job, checks that the company's VAT number is on it and settles each person's cash advance. A receipt takes about 4 minutes. Every month around 12 receipts are missing the VAT number, so the VAT can't be reclaimed, and nobody notices until three weeks later.
Now the technician takes a photo of the receipt on WhatsApp at the till and types the job name. The agent reads the store, amount, date and whether the company's VAT number is there. If it's missing, the reply comes in under a minute: "No company VAT number on this receipt. Ask for an invoice at the till." At month end, each person's expense claim is already filled in.
What stays with peoplechecking flagged receipts, matching against the paper originals and closing the advances.
Complexitymedium-high (photographed receipts, VAT rules, a claim per person)
Before: 24 h/month (300 x 4 min + 4 h claims) → After: 6 h/month
- Time recovered: 18 h x €33/h = €594
- VAT recovered: receipts without a VAT number drop from 12 to 3 a month; 9 receipts x €30 = ~€47 VAT (21%)
- Measured savings / month
- €641
- AIBT Labs fee (12%)
- €77
- You keep / month
- €564
Illustrative example
Incoming payments and overdue invoices
Commercial cleaning company, 180 clients on monthly contracts
Every morning someone downloads the bank statement and ticks off in Excel what has been paid. Sounds simple, but clients pay with references like "services", with no invoice number. Some pay three invoices in one transfer, others pay part of one. It takes about 50 minutes a day. On Mondays comes the overdue list, followed by emails and phone calls.
The agent matches payments by payer account, amount and invoice number where there is one. About 85% get matched on their own. The rest go to a person with a suggestion, such as "probably invoices 1042 + 1051". Five days after the due date, the agent drafts the reminder emails and a person approves the list before it goes out.
What stays with peopleunclear payments and phone calls to clients more than 30 days late.
Complexitymedium-low (matching on several criteria, unclear cases go to a person)
Before: 28 h/month → After: 10 h/month
- Time recovered: 18 h x €33/h
- Measured savings / month
- €594
- AIBT Labs fee (8%)
- €48
- You keep / month
- €546
Note: We didn't count the money that comes in sooner, even though that is usually the biggest gain.
Illustrative example
Reminders for unpaid invoices (email + WhatsApp)
Commercial cleaning company, 180 clients on monthly contracts, ~45 late payers a month
On Mondays the bookkeeper makes a list of overdue invoices. For each late client she finds the invoice and writes an email. If there's no answer, she writes again a week later. Reminders go out unevenly: some clients get three, others none. Now and then someone who paid the day before gets a reminder and is, fairly, annoyed.
The agent checks the bank statement daily and sees which invoices are past due and unpaid. It then prepares reminders in steps:
- +3 days: a polite email with the invoice attached
- +10 days: a WhatsApp message with the amount and IBAN, only to clients who agreed to be contacted on WhatsApp
- +20 days: an email with the account statement and all open invoices
- +30 days: the client goes on the call list
Each morning the bookkeeper sees the day's prepared messages and approves them. She can remove clients she's already talking to. When a payment arrives, reminders for that invoice stop on their own.
What stays with peoplecalls after 30 days, payment plan discussions and disputed accounts.
Complexitylow (fixed rules, the data is already in the bank statement and the invoice list)
Before: 18 h/month (4 h lists + 12 h emails + 2 h calls) → After: 6 h/month
- Time recovered: 12 h x €50/h = €600
- Capital collected sooner (assumption): 45 invoices x €300 = €13,500, collected on average 10 days earlier; cost of capital 12%/year -> 13,500 x 12% x 10/365 = ~€44
- Measured savings / month
- €644
- AIBT Labs fee (7%)
- €45
- You keep / month
- €599
Note: The time saved is small. The real gain is that every late client is contacted on time, every time. We didn't count it, but we show it separately on the dashboard, for example how invoices more than 30 days overdue go down month by month.
Illustrative example
Timesheets for field crews
40 workers, 5 crews, 8 job sites
In the evening, crew leaders send a WhatsApp message or a photo of the paper sheet, like "7 people at the Hill Street site today, Mark off sick". The next morning the office assistant types it all into Excel, per person and per site, and calls whoever forgot to send. At month end she builds the payroll timesheet and the hours per job for accounting. That's 1 h 15 min a day, plus a full day at month end.
Crew leaders send the same message as before, just to the company's WhatsApp number. The agent fills in the timesheet. At 7 pm it automatically messages anyone who hasn't reported. It also flags what doesn't add up: someone logged at two sites on the same day, or someone on leave showing up as working. The month-end timesheet builds itself.
What stays with peoplea 15-minute daily check and corrections before payroll.
Complexityhigh (free-text messages, cross-checks against leave, direct impact on pay)
Before: 34 h/month → After: 7 h/month
- Time recovered: 27 h x €30/h
- Measured savings / month
- €810
- AIBT Labs fee (15%)
- €122
- You keep / month
- €688
Illustrative example
Emails that always get the same answer
Building materials distributor, ~60 emails a day to the sales desk
When we timed it, about 25 emails a day fell into 4 types: "please resend the invoice", "when will my order arrive", "do you have the datasheet or declaration of performance for X" and "please send my account statement". Each one meant searching the ERP, finding the PDF, attaching it and replying: 5 minutes. Doesn't sound like much, but it adds up to 2 hours a day.
The agent recognises the type of request, pulls the data from the ERP and drafts the reply with the document attached. A person reads it and hits send, about a minute per email. For the first two months the agent only writes drafts. Automatic sending is switched on after that, and only for invoice resends.
What stays with peoplecomplaints, price negotiations and anything outside those 4 types.
Complexitymedium (recognising the request and connecting to the ERP)
Before: 44 h/month (25 x 5 min x 21 days) → After: 9 h/month
- Time recovered: 35 h x €33/h
- Measured savings / month
- €1,155
- AIBT Labs fee (10%)
- €116
- You keep / month
- €1,039
Illustrative figures. Each example is a single workflow, not the whole company. The fee depends on workflow complexity (5-15%) and is set after the audit. Hourly cost = fully loaded employer cost, based on the EU average (Eurostat).
05 / Projects
We started with our own companies.We test on real operations, then deliver.
We don't show up with slides. We first digitised our own group of companies: invoices, progress reports, subcontracting, field staff.
Construction and services / own group
Digitising our own group of companies
ProblemDocuments written by hand, accounting reports rebuilt every month, supplier quotes compared in spreadsheets and field staff scheduled over the phone.
SolutionAI agents and automations connected to the tools the group already used, introduced gradually and run in parallel until validated.
Automated workflows
- Automatic document generation: letters, notices, progress reports
- Automated accounting and reporting workflows
- Supplier quote analysis and procurement
- Resource and staff planning across projects
Measured resultMeasured figures will be published at the end of the measurement period.
- Let's measure your first workflow
02
The next case study could be yours
We publish every project in the same format: problem, solution, measured result. With the client’s consent.
06 / Lab
Labs isn't just a name.We test, measure and publish what we find.
We're practitioners, not resellers. Part of our time goes into building and testing, not just delivering.
- In daily useL-01
Internal tools
The tools we use every day in our own group: document generators, reporting agents, automatic checks.
- Lab notesL-02
We tested X on Y
Short technical write-ups: which model, on which workflow, with what result. Including what didn't work.
- In progressL-03
Experiments
New models, new automations, integrations with the systems SMEs actually use: accounting software, e-invoicing, WhatsApp.
- ReusableL-04
Impact study framework
The method we use to measure every workflow before and after: time, errors, cost. The same for every client.
07 / FAQ
Frequently asked questions.Straight answers, no fine print.
How much does it cost?
It depends on the process we automate. Our model: we measure the savings and you pay us 5-20% of them, on a 48-month contract. If you don't save, you don't pay. AI usage (tokens) is billed at cost, with no margin.
What happens to our data?
Your data stays yours. We sign an NDA before the audit. We do not store client data on our own servers and we work in line with GDPR.
How long does implementation take?
From 4 weeks for a simple workflow to 3 months for several processes. Every project starts with a pilot on the highest-impact workflow.
Do we have to change our existing systems?
No. We integrate with what you already use: accounting software, Excel, email, WhatsApp, your ERP. We don't replace, we tune.
What if employees don't want to use it?
We roll out gradually, with hands-on training. People adopt tools that make their work easier, not harder.
Does it work for small companies?
Yes. If you have at least one repetitive process that eats hours every week, we can help.
Do you work with companies outside Romania?
Yes. We work remotely with companies across Europe, with on-site visits for the audit when needed.
08 / Contact