Routine eats the day
Invoices checked by hand. Quotes rebuilt from old quotes. Enquiries lost after hours.
AI workshop · TRUSTBYTES Automator
AI that takes the office work off your desk. Safely.
We teach your team to use AI safely. Then we build an AI employee for the repetitive office work. A human always signs off.
The problem
Your best people are doing paperwork.
AI for the routine, a human for the decision
TRUSTBYTES’ Automator works inside the tools you already have. It books what it can prove and hands the rest to a person.
Step 1 · AI workshop
Your team learns AI in one session.
On site, for people from different departments. They carry the rollout inside your company.
- Length3–3.5 h
- Group5–7 people
What participants take away
- A shared vocabulary: model, token, agents, RAG
- A prompt template they used themselves
- A traffic-light rule for what may go into AI tools
- GDPR, works council and EU AI Act basics
- A first list of automation candidates
Workshop series
- Follow-up sessions at regular intervals
- Session two: we build a RAG together, a searchable knowledge base over your own documents
Step 2 · Automator
One AI employee. The same four steps for every job.
Pilot30 d
- 01 · Input
Something arrives: a voice note, an invoice, a bank transaction
- 02 · Read
The Automator pulls out the facts that matter
- 03 · Check
It compares them with your ERP, catalogue or open items
- 04 · Decide
Clear case: done automatically. Unclear: a person decides.
- Use case 1
Voice note → proposal
A voice note from the site visit becomes a draft proposal.
- Use case 2
Invoice automation
A supplier invoice is checked against the goods receipt and booked.
- Use case 3
Finance reconciliation
Bank transactions are matched to open items every day.
Step 2 · Use case 1 of 3 · Voice note to proposal
A voice note on site becomes a ready proposal.
−50 %less time per proposal
Estimate: a realistic range for today’s AI. We measure your real values in the pilot.
Today
- Notes on paper at the site
- Typed up in the evening
- Prices looked up by hand
- Sent days later
With TB
- Voice note on site
- AI drafts from your catalogue
- You check and adjust
- Sent within 1–2 days
- Time per proposal
60–90 min30–45 min - Time until the customer has it
3–5 days1–2 days - Proposals per person per week
~10~15
Step 2 · Use case 2 of 3 · Invoice automation
Supplier invoices checked and booked on their own.
~40 %of invoices booked without a human touch
Estimate: a realistic range for today’s AI. We measure your real values in the pilot.
Today
- Invoice opened by hand
- Typed into the ERP
- Checked line by line
- Mismatch found late
With TB
- Invoice read by AI
- Matched to goods receipt
- Clear match: booked
- Exception: to a person
- Handling time per invoice
10–15 min4–6 min - Invoices booked automatically
0 %30–50 % - From receipt to booking
5–10 days2–3 days
Step 2 · Use case 3 of 3 · Finance reconciliation
Bank and books reconciled daily, not monthly.
−60 %less time spent matching payments
Estimate: a realistic range for today’s AI. We measure your real values in the pilot.
Today
- Statement exported weekly
- Matched in spreadsheets
- Unclear items chased
- Close takes days
With TB
- Bank feed pulled daily
- Matched to open items
- Certain matches cleared
- Short list for accounting
- Manual matching per week
5–8 h2–3 h - Transactions matched automatically
0 %50–70 % - Month-end close
3–5 days2–3 days
The team
Built by a team backed by the US accelerator Techstars.
- Jeremy PrasetyoCEO & Co-Founder2x world champion. International keynote speaker.LinkedIn profile
- Marcelo Vieira FrayhaCTOLeads engineering from top to bottom. AI developer with 5+ years of experience.LinkedIn profile
- Jonas SurmannCo-FounderFounder of 3 companies. Master’s in AI and Blockchain.LinkedIn profile
The next step