Everything is referenced
Every answer comes with the document, the page and the clause behind it. If we cannot point at where something came from, it does not go in the output.
Our missionCritical work deserves
Critical work deserves
cutting-edge technology
Sourcestone was founded by actuaries and technologists who have spent their careers in UK pensions, financial data and responsible software. We started it because the most qualified people in a pensions practice were spending their weeks trawling through documents.
Why we started
There had to be a better way
Ask anyone who has produced a benefit specification what the job actually involves, and you get the same answer. Reading. A definitive deed, nine deeds of amendment, two consolidations, and an announcement from 1997 that quietly changed the escalation basis. Most of it scanned. None of it searchable.
It is skilled work, in that you have to know what you are looking for. It is not work that needs a qualified actuary to do all of it. That is the gap we build for.
Judgement stays with your actuary. It is the part clients pay for and the part that is hard. What we take away is the reading in front of it, and we show our working well enough that checking the output takes hours rather than weeks.
Nobody set out to leave the profession behind. Pensions software was built for administration, and benefits work quietly stayed manual because no one had a better idea. We think there is one now.
How we build it
Three things we hold to.
We flag, you decide
When two documents disagree, we describe the disagreement and show you both. We do not pick a winner. That call is yours, and so is the responsibility for it.
The awkward documents matter most
Scanned deeds with no text layer. Forty years of amendments. Member details buried in old correspondence. That is where the answers are, so that is what we handle.
What we are working on
Where we are and what's next
The engine reads documents and evidences answers. Benefit specifications and member data quality are what we have pointed it at so far.
- Benefit specifications, fully referenced, in your own template
- Member data quality mapped field by field, with sources
- Checking answers against standard checks, such as GMP positions
- Answering scheme-specific member questions from the documents
- Longer agentic workflows, carrying the same evidence trail end to end
This is what we are working on, not a set of dates.
Get started
See it on one of your own schemes.
Point us at one scheme's documents and see the referenced output on your own rules and your own template. Read access to one folder, revoked whenever you like.