Skip to content

Case study 001

AskIGD. Talk to ten thousand reports.

IGD's members pay for insight into the food and consumer goods industry, and that insight lived in more than 10,000 reports and articles. AskIGD lets a member ask a question and get a cited answer in about two seconds, with the sources one click away.

LiveLive internally since September 2026. Members-only, so ask us for a walkthrough.

Corpus
10k+reports and articles ingested, chunked and indexed
Speed
2.1saverage time to first answer
Delivery
6 wksfrom brief to first live, internally

01Context

An organisation that knows more than any one search can surface

IGD is a research and training body for the food and consumer goods industry, with members in more than 50 countries. Edge has run igd.com for two years.

The frontier report flagged the shift from search to conversation as the biggest change coming to how members consume research. The concept that followed was simple to say and hard to build: let people talk to the whole body of work, not browse it.

Where it started: the frontier report

02What we built

A conversational layer over everything IGD has published

Report ingestion at scale, with structure preserved so a chart or a table can be cited, not just a paragraph. Real-time factual augmentation from around the web when a question reaches past the archive, clearly separated from IGD's own findings.

Every answer carries citations with a snapshot of the source and a deep link into the IGD platform, so trust is checkable rather than assumed. And a new research tool for IGD's internal teams, who ask harder questions of the same corpus than members do.

A cited answer, with a snapshot of the source

03How

Multi-model, retrieval first, on Edge

Different models for ingestion, retrieval ranking and answer generation, chosen by measured performance on IGD's own material rather than by brand.

Retrieval is grounded in the archive before a model is allowed to write. Augmentation from the web is opt-in per answer, and labelled. The whole pipeline runs on Edge infrastructure alongside igd.com, so the answer and the source it points to live in the same place.

Retrieval first: the archive is searched before a model writes

04What changed

From “where is the report on X” to “what does IGD think about X”

Live internally in September 2026, after six weeks from brief to first working version and two months to the version now in use. IGD's own analysts use the internal tool to interrogate a decade of work before starting new research.

Before

A search, a download and twenty minutes of reading.

Now

2.1saverage time to first answer

A two-second answer, with the reading one click away if they want it.

05What's next

Members' agents will ask too

The next design decisions, covered in IGD's quarterly report, are about who is allowed to ask.

  1. 01Members' own AI agents querying research programmatically
  2. 02AskIGD as the front door for software as well as people
  3. 03Labs on the retainer, watching what comes next

Handed to Edge Expert Services to host, operate and evolve.

Engagement

Report, concept, build, handover.

The standard Labs path. Flagged in the frontier report, scoped as a concept with the number it would move, built by the Labs team, then handed to Expert Services to run.

AskIGD 001

Live
Client
IGD, a customer of Edge since 2024. Edge hosts and runs igd.com.
Sector
Food & grocery. Membership and research body.
Status
Live, internal launch September 2026. Members-only.
Build
Six weeks from brief to first live. Two months to the version in use.
Labs' role
  1. Frontier report
  2. Concept and spec
  3. Build
  4. Expert Services
Stack
Multi-model retrieval and generationWeb augmentation with source snapshotsEdge ComputeEdge StorageEdge CDN

Your archive has answers in it

If your organisation's value sits in a body of research, documentation or reporting that people have to hunt through, this is a conversation worth having.

Or write to labs@edge.network