Worked scenario · Manufacturing and after-sales
Four people answering the same question six times a day.
An 80-person manufacturer that builds and maintains professional equipment. 400 customers under contract, around 120 after-sales requests a week across three channels. Here is a full engagement, from the audit through to the agent that handles the requests.
The company
- Size
- 80 staff, including 4 in after-sales and 6 field engineers
- Business
- Manufacture and maintenance of professional equipment, 400 customers under contract
- Volume
- Around 120 requests a week: email, website form, phone
- Tools in place
- An ERP, a CMMS for interventions, a shared mailbox, a parts compatibility workbook, technical documentation in PDF
- What is missing
- No measurement of volume by reason. Nobody knows what actually occupies the team.
The starting point
After-sales receives around 120 requests a week. Nobody knows exactly what is in them, because nothing is categorised: they land in a shared mailbox, somebody reads them, somebody answers. Ask the manager what occupies the team and the answer is “a bit of everything”, and he is right.
Look at two weeks of messages and the reality is simpler: six reasons account for most of the volume. Where is my order, what is the lead time for a visit, is this part compatible with my model, am I still under warranty, where is the documentation for this equipment, can you schedule a visit. Six questions whose answer already exists somewhere in a system.
The cost is not the time spent answering. It is what does not get done during that time: the breakdown halting a production line waits behind forty order-status requests, and the first-response time is the same for both.
What the audit puts on the table
- 01
Across two weeks of analysed messages, six reasons cover 72% of the volume. The remaining 28% are cases that need human judgement and that, today, wait behind the rest.
- 02
Four of those six reasons already have an answer sitting in the ERP or the CMMS. It is simply not reachable without opening the tool and knowing where to look.
- 03
Parts compatibility lives in a workbook kept by one person. It is the most fragile point in the whole setup, and it is not an artificial intelligence problem.
- 04
Technical documentation exists as PDFs, with no structure and no clear versioning. Usable, but to be reworked before anything relies on it.
- 05
Average first-response time is 4 hours, and it is the same for an order-status request as for a blocking breakdown.
What gets built, week by week
- Weeks 1-2
Connectors and a written scope
ERP and CMMS connectors read-only, plus ingestion of the shared mailbox. In parallel, the agent's scope is written down in black and white with the after-sales manager: what it handles, what it does not, and who it hands over to in each case. That document is the real deliverable of these two weeks.
- Weeks 3-4
Warehouse and business model
A warehouse that names the objects of the business: customer, contract, installed equipment, order, intervention, part. The compatibility workbook is taken over, structured and versioned. Technical documentation is split up and indexed against its equipment reference.
- Week 5
The calculations, not the answers
Each of the six reasons is translated into a defined, validated query: order status, next scheduled intervention, warranty status, part compatibility, equipment documentation. The agent invents none of these answers, it picks the right query and returns it.
- Weeks 6-8
Progressive rollout
The agent starts in suggestion mode: it drafts the answer, a human approves in one click. After two weeks, the reasons whose approval rate exceeds 95% go autonomous. The others stay in suggestion mode for as long as needed. Nobody flips a switch and hopes it holds.
- Weeks 9-11
Escalation and measurement
Anything outside the scope is routed to the right person with the history of the exchange and what the agent has already checked. And for the first time, volume is measured by reason, by customer and by handling time: after-sales finally knows what occupies it.
What it changes
Share of volume handled end to end
None
60 to 70% after three months of running in
First-response time
4 hours on average, whatever the subject
Minutes on covered reasons, and complex cases move to the front
Team time on repetitive questions
Most of the day
What is left over. The team handles judgement calls and breakdowns.
Knowledge of volume by reason
None
Measured every week, by reason and by customer
Parts compatibility
A workbook, one person
Structured, versioned, queryable
These orders of magnitude are what this kind of engagement produces when the conditions below are met. They are not contractual commitments: the pricing for your own situation comes out of the audit.
What it costs, and how long it takes
Audit: €3,000, two weeks. Data foundation: €18,000, five weeks. Request-handling agent: €15,000, six weeks that partly overlap the previous ones. Around €36,000 over eleven weeks, decided in three steps. A monthly operating fee applies, covering on-call and the tracking of the escalation rate. You own the code and the credentials.
What would make this scenario fail at your company
Knowledge that only exists verbally
If parts compatibility is written nowhere and lives only in the heads of two engineers, it has to be written down first. That is doable, it takes time, and it is priced separately.
A scope nobody owns
The agent must be allowed to answer on its own for specific subjects. If every answer has to be approved indefinitely out of caution, you have added a step instead of removing one.
Aiming for 100%
The 28% of requests that need judgement should not be automated. Forcing them through produces confident, wrong answers, and that is exactly what loses a customer.
No measurement at the start
Without measuring volume by reason before starting, there is no way to say whether the setup works. That measurement is part of the audit, not of a dashboard added later.
The building blocks involved
The data audit
Two weeks at most, a fixed price, and a document that belongs to you. It is the only honest way to know what your data allows before committing a budget to it.
The data foundation
Four to eight weeks for your data to live in one place, historised and reliable. Delivered in usable stages, not in one go six months out.
The AI assistant
An assistant that answers your business questions in plain language, on figures that come out of your own database. It arrives once the foundation is reliable, never before.
The other scenarios
B2B trade and distribution
Monthly revenue lands twelve days after close, stockouts are discovered the day the customer calls, and four tools give four different definitions of revenue.
Hotels and restaurants
Occupancy, RevPAR and average spend live in three tools that don't talk to each other, and booking requests that arrive outside opening hours get lost.
Is your after-sales team still answering the same questions?
Two weeks of messages are enough to know what share of your volume can be automated, and what share must not be. That is what the audit does.