GEI-AI: What we are building | Digisarion Intelligence
GEI-AI scores the communication itself across three dimensions and returns a graded rating with named drivers. Currently in development.
We score the communication, not the audience.
Most approaches to this problem ask people what they thought after the fact. Surveys, focus groups, complaint volumes. All of them are slow and small, arriving after the communication has already gone out.
GEI-AI assesses the communication artefact itself before it is sent, across the whole population rather than a sample.
GEI-AI is in development. No product is commercially available, and no deployment has taken place. The descriptions on this page set out what is being built.
The three dimensions: Three independent measures, one rating.
Trust Score
How likely a communication is to be perceived as trustworthy by the person receiving it. This is the keystone measure and the headline output.
Emotional Resonance Index
The emotional response a communication is likely to evoke. A letter can be perfectly clear and still frighten the person reading it into not responding, which is a poor outcome by any regulatory standard.
Cultural Alignment Score
How well a communication aligns with the cultural norms and expectations of the audience it was written for. A tone that reads as direct and helpful to one group can read as abrupt or presumptuous to another.
The three combine into a single graded rating with the drivers behind it identified, so a communications team can act on the result rather than simply file it.
The output
A range, not a number. And always a reason.
Two design principles govern what GEI-AI returns.
Scores are expressed as a range with a stated confidence level.
A single decimal presented without an interval implies a precision that measurement of this kind does not have. Where the platform is uncertain, the output says so.
Every score carries its drivers.
A rating that cannot be explained is not usable by a compliance function, because the next question after "this scored poorly" is always "why, and what do we change?". An unexplainable score also cannot be defended to a regulator, which makes it worse than no score at all.
How it is intended to be used
One engine, three routes.
A dashboard for compliance teams.
Communications assessed across a reporting period, with coverage of the full population rather than a sample, and exportable documentation formatted for board reporting.
An interface for platform operators.
Verification at the point a communication is generated, intended for organisations facing transparency obligations under Article 50 of the EU AI Act.
Bespoke enterprise engagement.
For organisations with communication volumes, integration requirements or governance arrangements that do not fit a standard deployment.
What GEI-AI is not
The limits, stated up front.
It is not an audit, and does not certify
GEI-AI is an assurance tool. It strengthens the evidence an organisation produces about its own communications. It does not issue a compliance certificate, and it does not substitute for the organisation's own judgement or its regulator's assessment.
It does not replace human review
It makes human review possible at a scale that manual reading cannot reach, and it directs attention to the communications that need it.
It does not assess people
The unit of analysis is the communication. GEI-AI does not profile, score or make inferences about individual recipients.