Same working day
from 4–6 weeks
Research file to usable reporting
Evidence a Funder Can Trust, Produced In-House
How London Youth Rowing freed its partnership reporting from specialist consultants, with AI agents built by AgentCo on AWS
Customer
Industry
Sports & Community Development
Location
London, United Kingdom
Partner
Experienz Limited
Background
Executive Summary
London Youth Rowing (LYR) is a UK national charity that uses rowing to widen opportunity for young people from underrepresented and disadvantaged communities. Its Active Row programme runs in more than 100 state secondary schools across London, Yorkshire, Nottingham, Kent, Bristol and the Thames Valley. Corporate partnership has become an important part of how that work is funded, and partners increasingly expect evidence: who was reached, whether awareness of their involvement actually moved, and what their investment delivered.
To answer those questions LYR commissions audience and brand research. The problem was that the research arrived in a form nobody at LYR could use. Raw survey files needed statistical preparation before any trend could be read, and LYR has no internal data team. Every research wave meant paying for specialist help, and every new report meant waiting on an external analyst. Evidence that should have supported a live partnership conversation routinely arrived after it had happened.
AgentCo replaced that process with two AI agent services running on AWS. The first prepares each research wave automatically: it reads the raw survey file, applies the statistical treatment the research design requires, and withholds any figure that rests on too few responses to be reliable. The second lets a member of the LYR team build a partner-ready dashboard simply by describing what they want to see.
As a result, LYR can now produce credible partnership evidence on its own timetable, without commissioning specialist analysts or hiring a data team, and can repeat it every research cycle at no additional preparation cost.
The Challenge
The Business Challenge
Commissioning the research was never the difficulty. Using it was.
The research arrived as raw data, not as answers
Survey results are delivered as raw statistical files, one per wave. Before anyone can read a trend from them, three things have to happen, and none is a matter of opening a spreadsheet.
Mistakes did not announce themselves
The most uncomfortable part of the old process was that errors were silent. A survey file mishandled during preparation does not produce an obvious failure. It produces a believable number that happens to be wrong, and it stays wrong until someone outside the charity notices. For an organisation whose partnership case rests on the credibility of its evidence, that is a reputational exposure as much as a data problem.
Why this was a problem for LYR specifically
What LYR Needed to Achieve
Working with AgentCo, LYR set out four requirements. All of them were business requirements rather than technical ones.
Turn each research wave into trustworthy trend evidence without commissioning a specialist to do it.
Produce partner-ready views on demand, in-house, by people who are not analysts.
Guarantee that unreliable figures can never be published, rather than relying on someone to catch them in review.
Repeat all of the above every cycle at no additional preparation cost, so that adding partners does not add overhead.
How AgentCo Solved It
AgentCo built two AI agent services on the AgentCo platform, hosted on AWS. Together they replace the two activities LYR was previously buying in: preparing the data, and building the reporting.
Automatic preparation of every research wave
When a research wave arrives, it is handed to the platform rather than to a consultant. A team of AI agents takes it through four steps without human intervention
When a research wave arrives, it is handed to the platform rather than to a consultant. A team of AI agents takes it through four steps without human intervention
The important part is that this happens as the data is prepared, not as a review afterwards. By the time anyone at LYR looks at a number, the statistical protections have already been applied to every possible view of it. Reliability is a property of the dataset rather than something a reviewer has to remember to check.
Reporting built by asking for it
The second service replaces the external BI contractor. A member of the LYR team describes the report they want in ordinary language, for example:
“Build an awareness tracker: trend by wave, a headline KPI, a metric breakdown.”
An AI agent then does the work a BI analyst would have done. It establishes which data the person is allowed to see, looks at what is genuinely available for that programme or partner, checks the real numbers and sample sizes behind the request, designs the dashboard around what the data will actually support, reviews its own draft for problems, and produces a finished, partner-safe dashboard. The user watches it happen and has a working report in minutes rather than weeks.
Because the agent works out what is available each time it runs, the same service works for every programme, event and partner without anything being rebuilt for each one.
Why this needed AI agents rather than a script
Conventional automation works when every input looks the same. Survey files do not. Questions get reworded between waves, answer scales get revised, new demographic breakdowns get added, and different research suppliers structure their files differently. A script written for the last wave breaks on the next one, which is precisely why the work kept going back to a specialist.
The agents examine each file and each request and decide how to handle what is actually in front of them, rather than following instructions written in advance for a file nobody has seen yet. When something does not add up, such as an empty result, a breakdown that does not reconcile, or a segment too small to report on, they detect it and correct it before anyone sees the output. That self-correction is what removed the specialist from the process rather than merely making the specialist faster.
It is also why a conversational chatbot would not have solved this. A chatbot can answer questions about data that is already trustworthy. LYR needed something that makes the data trustworthy in the first place, and then builds the reporting on top of it.
Keeping partner data separate and safe
LYR is configured as an isolated tenant on the AgentCo platform, and every report is bound to the security context of the person requesting it. Access is established by the user’s session rather than by anything they type, so a request cannot widen its own reach. The reporting agents can read data but can never change it, and the underlying research records are never altered by the reporting process. Given that the wider charity handles data relating to young people, this separation was a condition of the design rather than a feature of it.
What Changed
The Benefit to London Youth Rowing
The value of the solution is best measured against what LYR exists to do, which is deliver programmes for young people rather than administer research.
Money returned to the mission
Producing credible partnership evidence previously required extended specialist consultancy effort every cycle. That spend has been removed from the routine reporting process. For a charity, analysis cost avoided is programme funding retained.
Independence from external suppliers
LYR no longer depends on an analyst’s availability to answer a partner’s question. The team can produce the view themselves, during the conversation that prompted it, and can revise it as the discussion develops.
A stronger partnership case
Partners and funders can be shown live, interactive evidence rather than a static report produced weeks earlier. Every figure carries its own provenance, and unreliable figures are structurally prevented from appearing at all, which makes the evidence easier to defend under scrutiny.
Capacity to grow partnerships without growing overheads
Because preparation and reporting no longer repeat manually for each wave and each partner, the reporting effort no longer scales with the size of LYR’s partnership portfolio. Winning more partners no longer creates a proportionate reporting burden.
Measurable outcomes
The measures below were agreed as the basis for evaluating the solution, and were recorded over the first two full research cycles following go-live.
The Measurable Difference
London Youth Rowing | First Two Research Cycles After Go-Live
Conclusion
London Youth Rowing needed to prove the value of its partnerships without building a data team it could not afford, and without paying specialists every cycle to prepare data it had already paid to collect. That was a fundraising problem before it was a technology one.
The two agent services AgentCo built removed the bottlenecks that stood in the way. Research waves are prepared automatically and safely the moment they arrive, and reports are built by the people who need them, at the point they need them. The charity now owns its evidence from end to end.
The value did not come from the sophistication of the technology. It came from putting a capability that had always sat with external specialists into the hands of a small charity team, and doing it in a way that makes an unreliable figure structurally difficult to publish. For LYR, that means more funding and more attention pointed at Active Row rather than at administering research, and a stronger case to put in front of the next partner.