Route AIAdvisory
References

From practice

A selection of recent work. We don’t name our clients; on request we are happy to put you in touch with a reference.

Training and e-learning

Copilot training and e-learning for an engineering company

Copilot agents

Copilot agents for client memos at an international service provider

Dossier analysis

Complex insurance case: full report within 15 minutes

Dossier analysis

Freedom of information file: from released documents to a readable story

Training and e-learning

Copilot training and e-learning for an engineering company

Client: Machine builder and steel construction company in the west of the Netherlands, with office, engineering and workshop

The challenge

The company wanted its employees to learn to work with Microsoft Copilot. Shortly before the training, the group turned out to be mixed: some had only Microsoft 365 Business Basic with Copilot Chat, others a full Copilot licence.

Approach

  • Training in two 2-hour sessions with a practice period of four to six weeks in between
  • One programme for both groups: session 1 in the chat for everyone, session 2 on what a licence adds, with a variant of every exercise for both situations
  • Participants worked in pairs, always one with and one without a licence
  • Practice files from their own sector with fictitious data: an order overview, a project file and sample emails
  • Adoption treated as a behavioural question: knowing why, wanting to, being able to and keeping it up
2×2hours of training plus practice period
12hands-on exercises
60e-learning screens for reference

Delivered

  • Training deck of 58 slides with twelve exercises and extensive trainer notes
  • E-learning for reference, one module per session, protected with an access code
  • A licence label per topic, so every participant sees what works for them
  • Short knowledge checks with an explanation after each answer

What we learned

Don’t measure the number of prompts, measure how many people have three recurring AI tasks. And never present a licence gap as a shortfall, but as an insight: which jobs justify a more expensive licence?

Phase 1: Start
Copilot agents

Copilot agents for client memos at an international service provider

Client: International service provider in the insurance sector

The challenge

Relationship managers spent a lot of time drafting client memos. Quality and structure varied from person to person.

Approach

  • A framework of Microsoft Copilot agents: one dedicated agent per client, drawing on that client’s folder
  • Shared instructions, standard prompts and a fixed memo template, managed centrally
  • Phased roll-out: baseline, pilot with one client, live demo, feedback rounds and roll-out
  • Processing inside their own Microsoft 365 environment, with a DPIA before go-live
1agent per client
30%targeted time saving per memo
8 wksfrom pilot to roll-out

Delivered

  • Agent architecture with data isolation per client
  • Manageable instruction set and memo template
  • Roll-out plan and measurement set-up

What we learned

One agent per client provides data isolation and focused context, and makes scaling straightforward.

Phase 2: Apply
Dossier analysis

Complex insurance case: full report within 15 minutes

Client: Party in insurance proceedings before the Dutch financial services complaints institute (Kifid)

The challenge

The key question was whether three claims were legally connected, with a single payout, or not, with three payouts. The file contained policies, case law, correspondence and pleadings.

Approach

  • Documents made readable and sorted into five categories
  • Cited case law identifiers (ECLI) validated
  • Signals for and against connection visualised, with a source reference for each finding
  • Processed entirely locally
16source documents
<15 minfrom start to report
0documents sent to the cloud

Delivered

  • Report with timeline, core summary and points of attention
  • Source reference for every conclusion

What we learned

AI does not replace a lawyer, but makes sure you know your file before expert hours start to count.

Dossier analysis
Dossier analysis

Freedom of information file: from released documents to a readable story

Client: Requester under the Dutch Open Government Act (Woo)

The challenge

A freedom of information request produced thousands of pages of partly redacted, unsorted documents and emails.

Approach

  • Scans and email exports converted into searchable text
  • Duplicate documents and attachments filtered out
  • Timeline and thematic structure: who knew what, and when

Delivered

  • Thematic analysis with timeline
  • Overview of what is missing or redacted, with source references

What we learned

Deduplicating before analysing saves time and cost.

Dossier analysis

A similar question?

Tell us where you stand. We are happy to think along about an approach that suits your organisation.

Start the AI check →Book a call