How do you cleanse and enrich 138,613 alumni records without draining fundraising time and budget?
Newcastle University's fundraising team works from a CRM of around 287,000 alumni records gathered over decades. The data was the foundation of every fundraising conversation, but keeping it current was close to impossible. Database managers patched records by hand, and a single prospect record took about 30 minutes to work through. Fundraisers were losing close to a fifth of their time to verifying and cleaning records before they could pick up the phone.
Almost half the database, 138,613 records, sat outside the prospect funnel entirely. Working through that backlog by hand would have taken roughly 70,000 people-hours, near enough eight years of continuous work, at a cost of about £1.4 million.
We've built an agentic enrichment pipeline that works through alumni records one at a time, end to end, with no researcher in the loop. It runs in three stages.
Fetch
A cohort is pulled from Raiser's Edge using a saved query, for example every cold, disqualified contact. The agent picks records off that list one by one.
Research
For each record, the agent reads what the CRM holds, then looks for the same person across public and third-party sources. It searches the open web and LinkedIn through Tavily, finds work emails through Full Enrich, and checks the UK director registry through Companies House.
A GPT-5.1 model weighs each result against the original record and returns a match decision with a confidence score, so the agent knows it has the right person before trusting anything. Where a direct match fails, it falls back to the director lookup. Where a full match is not possible, it writes partial enrichment.
Commit
Verified fields are written back to dedicated custom fields in the CRM, and the agent saves its identity-match reasoning against every record, so any decision can be checked later.
Increase in contactable alumni,*
across the records processed so far.
The number of alumni with a verified email on file rose from 8,167 to 19,857. That is 11,690 more people the team can now reach.
The agent has processed 61,215 of the 138,613 records, around 44%, for about £1,879 so far.
On completion, the whole backlog will have been enriched at a pace and cost no manual process could match. The eight years and £1.4 million it would have taken by hand collapse into a single automated pipeline.
Beyond the numbers, the work has changed how the team operates. The data is something they can act on with confidence. Staff who were cautious about AI at the start are now engaged and curious, and Newcastle has gone from no in-house AI experience to running a sector-leading system.
*Measured across the 61,215 records (44% of the 138,613 cohort) processed between April and June 2026. 'Contactable' means a verified email address is on file.
