Manufacturing automation without replacing your ERP
The production data is already yours: in the ERP, in shift spreadsheets, in email and on paper by the machine. We pull it into one place and give you an AI agent that answers questions about output, scrap and stock the moment you ask them, instead of three days after the shift.
- 14 daysfrom the end of the workshop to the first shift report that builds itself
- EUR 3,500net per week of workshop, a published price we do not negotiate
- 3 to 5 timeshow much more a repair after a breakdown costs than a planned inspection
The problem
Why is the board looking at production from three days ago?
Machine cycle time, raw material consumption, scrap rate, unplanned downtime, output per shift, work in progress. Every one of those numbers exists in your plant. The trouble is that each one lives in a different system than the one next to it, and some of them an operator writes down by hand and hands over to the next shift.
Planning a multi-product line on shared resources is an optimisation problem. A spreadsheet solves it a week late, with no allowance for breakdowns or absence, so the plan and the shop floor part company on the second day of the week.
Quality control at the end of the line catches a defect once the batch is already made. An automotive customer expects full batch traceability under IATF 16949 at the same time, and rebuilding it out of four spreadsheets costs a process engineer a day.
- The board report is assembled by hand, so Monday's meeting is about last Thursday
- Too much of one component and none of another in the same week, because stock is tracked from memory
- Maintenance reacts after a breakdown instead of before it, because nobody sees machine parameters as a trend
- One shift supervisor leaves and the knowledge of how the line is really set up leaves with them
Why now
Why does this pay off today when it did not three years ago?
Three years ago, wiring an ERP to a warehouse system and to shop floor data was a quarter-long project on a budget that closed the subject for any small or mid-sized plant. Today the connection runs through interfaces that already exist. Collection goes through n8n or Make, the data lands in BigQuery, in a SQL database or in a spreadsheet, and the cost is counted in days of work rather than months.
What can be automated changed as well. A rule will move a number from one spreadsheet to another, but it will not answer the question of why scrap doubled on the second shift yesterday. A language model reads data out of several systems at once and answers in a sentence, rather than with a table somebody still has to work through.
Then there is the shortage of experienced planners and shift supervisors. A plant that has nobody to hire for that role has two options: overload the people it already has, or give them a tool that takes the data collection off them and leaves them the decision.
What changes
What is visible after the first month?
The shift report builds itself
Data arrives from the ERP, from spreadsheets and from the shop floor with no person in the middle. The supervisor does not assemble the report, they comment on it.
3 to 5 hours a week, the time it takes to stitch a production report together by hand
A deviation shows up the same day
The alert fires the moment scrap, output or material consumption drifts from the norm. You react on the shift where it happened.
An alert instead of a weekly review after the fact
Batch traceability in one question
Where the material came from, who ran the shift, what the parameters were and where the batch went. An answer instead of a day of a process engineer's time.
IATF 16949 and ISO 9001 requirements handled by a query, not by rebuilding from spreadsheets
Stock counted, not remembered
Levels, consumption and supplier lead times in one place, so you can see which component runs out in two weeks.
No more surplus on some part numbers while others are missing
Maintenance before the breakdown
Machine parameters collected as a trend, not read off at the moment of failure. The inspection gets planned into a stoppage, not into the middle of an order.
A repair after a breakdown costs 3 to 5 times more than a planned inspection
Proof
Do you use this yourselves?
Yes, and it is the only proof we can show without asking a client for permission to publish their figures. Two AI agents running in production inside our founders' own companies, both launched in March 2026 and built by the same method described above. Below is six months of measurement, together with what these numbers do NOT prove.
Built for ourselves. These are not client deployments but implementations inside the founders' own companies. We say so plainly, because proof you cannot verify is not proof.
JustBottheir agentJustIdea Agency, a performance marketing agency of around 30 people
- 23
- people from the company work with the agent
- 38
- systems wired in permanently
- 112
- processes run without a human
- 475
- thousand lines of the agent’s code
An agent wired into Google Ads, Meta, the CRM, the task system, the SEO tool, Analytics, Search Console and the data warehouse. First commit 1 March 2026, today 23 people use it.
8 of 38 systems it works in
Google Ads
Meta Ads- HubSpot
- ActiveCollab
- Ahrefs
GA4
Search Console
BigQuery
- Before
- In the twelve months before launch the agency ran at a margin of minus 4.42 percent. Across a twenty-month window there was not a single run of three profitable months.
- After
- May, June and July 2026 are three profitable months in a row, the first such run in the whole window. Revenue barely moved, so the change sits on the cost of delivery, not on sales.
Margin up 5.55 percentage points year on year, with revenue up 1.5 percent
SurfBottheir agentSurf Inc., a premium clothing brand with an online shop and two physical stores
- 16
- people from the company work with the agent
- 21
- systems wired in permanently
- 89
- processes run without a human
- 323
- thousand lines of the agent’s code
An agent wired into the order system, the ERP, the shop, the data warehouse and the ad platforms. Launched in March 2026, fully running by May. Sixteen people use it, warehouse and store staff included.
8 of 21 systems it works in
PrestaShop- BaseLinker
- Comarch Optima
- BigQuery
- Google Ads
- Meta Ads
- GA4
Zalando
- Before
- From February to August 2025 every unit of ad spend came back as 3.53 in online revenue, and advertising absorbed 28.3 percent of that revenue.
- After
- In the same months of 2026 the ratio is 4.19 and advertising is down to 23.9 percent of revenue. The ad budget is 6.9 percent lower with sales 10.3 percent higher.
Return on ad spend from 3.53 to 4.19, 18.5 percent higher, on a budget 6.9 percent lower
2026, with the agent 2025, before the agent revenue per unit of spend
Return on every unit of ad spend at the clothing brand, February to August, 2026 against 2025. The 2026 curve starts LOWER than the year before, breaks away in March and holds the lead for the rest of the season. The peak is 4.91 in May. We point out the weaker start because it shows we did not pick a convenient baseline.
Measurement: Online revenue from the order system against spend across every ad platform, the same months in both years. Cost data in the warehouse starts in February 2025, so we cannot rule out that the ratio was already rising on its own.
Operations run by the agency agent in successive months of 2026, from the first commit in March. From 83 operations in March to almost 63 thousand in August, with users growing from 4 to 23. This curve is here because a deployment nobody uses looks identical in a report to a successful one.
Measurement: Agent call log in the data warehouse, March to August 2026, counted at the level of a single tool invocation.
Breakdown of 29,854 operations run without a human over thirty days, by kind of task. Most of it goes to tracking team availability and monitoring client sites: things that have to run without a break and that nobody wants to do by hand.
Measurement: Agent audit log, the 30 days before 2 September 2026, counted by the kind of tool invoked.
Of 64,903 operations run in the last thirty days, 29,854 went through without a human: client site monitoring, warehouse loading, ad account reviews, catching form submissions. The rest is work someone requested and reviewed. We give both numbers, because autonomy without oversight is not a feature but a risk.
Measurement: Agent audit log, the 30 days before 2 September 2026, split by whether a call had an associated human session.
And in your plant
What if we are different?
Four sentences we hear on the shop floor more often than any others. All four are fair, so we answer them the way we would on a call.
We have an old ERP and we cannot replace it.
You do not have to. We read data out of it, we do not stand in for it. Swapping a production system is a year-long project, and this is fourteen days to the first working report. Systems with no modern interface are handled through a database read or a scheduled export, and we settle which in the workshop.
Operators write some of the data down by hand.
That is the normal starting point and it does not block the start. We begin with what is already in the systems, and we turn a handwritten note into a form only where that number actually feeds a decision.
We have no IT department, nobody to look after it.
The technical side is on us. What stays after the rollout is run by a person who knows production, not by a developer. Training that person is part of the work, not an extra.
Production data is our trade secret.
Which is why we work on your infrastructure or on a cloud you nominate, and the scope of access is agreed in the workshop and written down. We sign a data processing agreement under the <a href="https://eur-lex.europa.eu/eli/reg/2016/679/oj" rel="noopener">GDPR</a> before we touch anything. Nothing leaves the scope that was signed.
- ERP systems
- SAP Business One, Microsoft Dynamics 365 Business Central, Sage, Odoo, Infor, Epicor
- Production and quality
- MES systems, shift spreadsheets, control charts, machine data over OPC UA
- Warehouse and purchasing
- WMS, inventory spreadsheets, supplier portals, barcode and RFID scanners
- Teamwork
- Microsoft 365, Google Workspace, Slack, task systems
- Data and reporting
- BigQuery, SQL databases, Power BI, Google Sheets
- Automation layer
- Make, n8n, custom API integrations
What exactly we automate in a plant
| Area | Today | After the rollout |
|---|---|---|
| Shift report | the supervisor assembles it by hand, 3 to 5 hours a week | ready at the end of the shift, the supervisor comments on it |
| Scrap and waste | visible in the weekly summary | an alert the same day, as soon as the threshold is crossed |
| Batch traceability | rebuilt out of spreadsheets, around a day of work | an answer to a question asked in plain language |
| Component stock | tracked from memory and from a spreadsheet | a shortage forecast from consumption and supplier lead times |
| Maintenance | a reaction after the breakdown | a parameter trend and an inspection planned into a stoppage |
| Request for quotation | retyped between email, a spreadsheet and the ERP | entered once, with an audit trail behind it |
The scope of the first rollout is agreed in the workshop. We take the three items with the highest return, not all of them at once.
How we do it
Four steps from a spreadsheet to an AI agent
Workshop on the shop floor
One week. We walk the process where it happens: from the order, through planning and the shift, to quality release and dispatch. We come out with a map and a list of the places where data goes missing.
Picking three automations
From the map we take the three with the highest return and we measure them before we start, so that after thirty days it is clear whether they worked.
The first rollout
Fourteen days to a working solution. We usually start with the shift report, because it takes the most hours off people and shows the real state of the data fastest.
Measurement, then the next one
After thirty and ninety days we check the indicators agreed in the workshop. Only then do we decide on the next area.
Before and after
How a shift supervisor's week changes
Before
- An hour a day copying shift data into a spreadsheet
- The scrap summary lands on Friday, once the batch is already with the customer
- A call to the process engineer to rebuild the history of a batch behind a complaint
- The plan parts company with the shop floor on Tuesday and nobody sees it until Thursday
After
- The report is ready at the end of the shift, to be read and commented on
- A deviation alert the moment it happens, while the batch is still standing
- Batch history as the answer to a question asked in an ordinary sentence
- A drift from plan visible on the day it appears
What changes is when the information arrives, not how much of it there is. The data was all there before.
For the owner and the production director
What the person signing this off gets
The conversation about production stops being a reconstruction of last week. The meeting shows the state as of today, so the question becomes what to do about it rather than where that number came from.
The second thing is less obvious: the knowledge of how the line really runs stops living only in two people's heads. That takes away a risk nobody in a plant likes to name out loud, until somebody hands in their notice.
What it costs
What this costs
The process workshop costs EUR 3,500 net per week and we publish that price, because a plant has a right to know what it is agreeing to before it picks up the phone. In manufacturing it usually runs one week, and up to seven when several lines or several sites are in scope. We quote the rollout after the workshop, once we know how many systems we are connecting and what state the data in them is in.
Tool cost goes to the vendor directly, with no margin from us. The AI model cost is a vendor cost as well, in the range of EUR 0.005 to 0.07 per query depending on the model and the length of the answer.
Process workshop
EUR 3,500 net
per week, one week for a single plant
A plant that wants to know where it actually loses time before it buys anything
- A week on the shop floor and in the systems
- A process map from the order to dispatch
- Three automations with the highest return, costed before the start
- The indicators to measure after 30 and 90 days
- Most popular
Workshop and first rollout
Quoted after the workshop
scope follows from the number of systems and the state of the data
A plant that wants a working shift report inside two weeks
- Everything in the workshop
- The first automation live 14 days after the workshop
- A connection to the ERP and to the spreadsheets, with no system replaced
- Training for the person who runs it on your side
- Measurement after 30 and 90 days
AI agent for production
Quoted after the workshop
for plants where a dozen people a day ask about production
A plant where questions about output, scrap and stock come in all day
- Everything in the first rollout
- An agent answering questions about output, scrap, stock and batches
- Deviation alerts with thresholds agreed in the workshop
- Access for shift supervisors, planners and maintenance
If after the workshop we think there is nothing worth automating, we say so and we do not propose a rollout. It has happened and it will happen again.
Frequently asked questions
How is production automation different from an MES?
An MES controls how production is executed and it is a system you roll out over years. What we build sits on top of what you already have: it reads data from the ERP, from spreadsheets and from the shop floor, and answers questions about it. A plant with an MES gains as much as a plant without one, because the problem is rarely missing data and almost always data sitting in four places.
Do we have to replace the ERP?
No. We read data out of it through an interface and leave it where it is. Replacing a production system is a year-long project, and the first working report is meant to exist in fourteen days. Where a system has no modern interface, we take a database read or a scheduled export instead, and we settle which one in the workshop.
How long from signing to the first result?
One week of workshop, then fourteen days to the first rollout counted from the end of it. We measure the effect after thirty and ninety days against indicators agreed before the start, so that nobody has to judge it on impressions.
What if operators write data on paper?
That is a normal starting point. We begin with the data already in the systems, because it is enough for the first report. A handwritten note becomes a form only where that number actually feeds a decision, not everywhere at once, and the plant decides where that line runs.
Will this cover IATF 16949 and ISO 9001 requirements?
Batch traceability, meaning material origin, process parameters, the shift that ran it and the customer it went to, comes back as a single answer to a single question. It does not stand in for an audit or for your quality system documentation, but it takes a day of a process engineer's work out of every complaint.
What does the rollout cost after the workshop?
It depends on how many systems we connect and on how ordered the data in them is. We quote after the workshop, because that is when we know the answer. The workshop itself costs EUR 3,500 net per week, and you know that price before you call us.
Let us start with one question
The free consultation takes 30 minutes. We walk one of your processes and tell you whether we see anything in it worth automating. If we do not see anything, we say so.
We run two workshops a month, because a week on the shop floor is a week of one person's time on our side.
Book a free consultationPS. If you want to see what such a workshop looks like before you book anything, we describe it step by step on the product page.