Promtech SystemsData for decisions
You know your industry. We dig out the data your decisions need, from your own systems and from the world outside, join it up and put it where the decision gets made. The conclusions stay yours.
SituationHow it works today
StanceWhy us
Digging out data for the people who decide is what we have done for fifteen years. We don't know your industry better than you do. But we can help you know more
SourcesThree pipes
Source.01Internal
The numbers already sit in your ad platforms, CRM, finance system and spreadsheets. We dig them out, make them agree with each other and put them where you already are.
Seen in · P&L system · Fifty / Analytics in Telegram / Nordic reporting · Ve / Sales enablement suite / Analytics · Semetrix
Ads · CRM · Finance · Sheets → one view, every morning
Source.02External
What shows on retailers' shelves, in prices and in public sources, collected every day and presented as facts and ranges, each with a date. What it means for your business is your call.
Seen in · ShelfScore
Shelves · Prices · Public sources → dated facts, every day
Source.03Both, joined
The most useful answers sit where the two meet: which companies watched your ad, next to your own outreach. What the traffic was worth, next to what it cost.
Seen in · AudienceScan / Reach Ledger · Sentora
Your outreach + who watched → who to call, ranked daily
ApproachThree phases
Every engagement moves the work from our hands to yours. By the end of phase three you run it without us.
Phase 01 · Week 1
One bounded decision, not an AI strategy. We sit with the people who make it and ask what they did last time, which data was missing and what that cost. You say what better means, in hours or kronor.
Phase 02 · Weeks 2–4
The data is pulled out, joined and tested, then lands where the decision is made: the chat, the inbox or one simple view. A working system in your environment, not a demo.
Phase 03 · Week 4 →
You own the code, the tests and the docs. No subscription, no consultant dependency. We measure the after, then move to the next use case.
Your cloud
Deployed into your own cloud project when you want it there. Google Cloud is home turf.
Your data
Built on the warehouse you already have: BigQuery, Postgres or Supabase.
Your tools
Output lands where the team already works: the CRM, the inbox, a daily dashboard.
Toolboxpromtech stack
The same proven components under every system, so a pilot starts from working parts instead of a blank page.
01
BigQuery and Postgres, modelled in dbt, with tests that fail bad data before anyone sees it.
02
LLM agents with explicit context, memory and tool use. Claude first.
03
MCP tool definitions generated from an API's own OpenAPI spec.
04
Evaluation and failure-mode analysis treated as first-class work, not an afterthought.
05
Budgets, caps and circuit breakers enforced in code, not remembered.
06
Schema-aware context, so agents answer plain-language questions from your warehouse.
07
HubSpot, LinkedIn and the ad platforms, including the parts they don't document.
08
A Monday summary, or an alert when a number moves more than it should.
09
A health record for every run, so silence never passes for success.
10
Answers where people already are: the chat, the inbox, the CRM. Dashboards only where they earn their place.
WorkSelected cases
Systems built for teams at: Fifty Technology · Ve Interactive · Sentora · Semetrix Digital · Interact Works
Measurednot estimated
Start with one person and one task. If the number isn't worth it, we shouldn't build it.
spends hours a week on .
Erik earns SEK a month.
Based on 47 working weeks and full-time hours (160 h a month).
Tell us about the taskFAQStraight answers
Promtech digs out the data a decision needs, from your own systems, from the world outside or both joined, and delivers it where the decision gets made: the chat, the inbox or one simple view. You keep the industry knowledge and the conclusions; we build the pipes that bring the numbers to you.
Teams that make the same decision every week or month on numbers that arrive late or don't agree with each other. Past examples: a trading desk deciding where media budget goes, brands that sell through retailers, and a B2B marketing team choosing which companies sales should call.
No, and we don't pretend to. You know your industry; we know how to get data out of systems, join it with what happens outside and keep it honest. We ask a lot of questions, often because we don't know the answer, and that is where the gaps usually show up.
Three kinds: your own systems (ad platforms, CRM, finance system, spreadsheets), the world outside (retailer shelves, prices and public sources, collected every day) and the two joined. Every number carries the date of its data, measured facts are shown as facts and estimates as ranges.
Three phases over about four weeks. Week 1: we pick one bounded decision, sit with the people who make it and agree what better means, in hours or kronor. Weeks 2–4: the data is pulled out, joined, tested and delivered as a working system in your environment. From week 4: we hand it over and measure the result.
You do: the code, the tests and the documentation. It runs in your own cloud project and on your own data warehouse (BigQuery, Postgres or Supabase) when you want it there. No subscription and no consultant dependency.
Where your people already work: WhatsApp, Slack, Teams, Telegram, email, the CRM or one simple daily view. People can ask a question in plain language and get an answer from your own data, with a short summary before Monday's first meeting. Dashboards only where they earn their place.
Access is limited to a short list of approved people; anyone else gets no answer, and every question is logged. The system runs in your environment and only sees your own data, and tests stop bad data before anyone sees it.
It depends on the decision and the data, so we start by measuring what the problem costs today. If that isn't worth fixing, we shouldn't build anything. The calculator above gives a first number for one repetitive task.
In Stockholm, working in Swedish and English. The company is Promtech Systems Ltd, registered in the UK (company no. 13090657), and is led by Niklas Andersson.
Send a few lines to Niklas@promtech.systems: which decision, who makes it and what's missing. We reply within a couple of days.
You know your trade.We know the pipes.
A few lines are enough: which decision, who makes it and what's missing. We reply within a couple of days.