Ideas into
living worlds.

The brief

Give every user an intelligent health companion that interprets their biomarkers and symptoms the way a clinician would, not the way a generic wellness app does. Shift healthcare from reactive and fragmented toward proactive and continuous.

AverCare arrived with a concept and not much else. No brand, no character, no platform, which suited us, because that is the whole job: design the avatar, build it, and build the place it lives.

The founder had taken the idea to other technology partners first. None of them could build it.

The Challenge

Healthcare and AI is one of the most constrained environments you can build in. The product had to interpret medical data credibly without overstepping into diagnosis, ingest continuous data from wearables that all speak different formats, and satisfy compliance review before a single user touched it.

Each of those pulls against the others. Too much caution makes the companion useless; too much confidence makes it dangerous. And Aver had to be warm enough that people would actually talk to her, without ever sounding more certain than the evidence allowed.

What we built

• The brand. Name, identity, visual language and tone of voice, built before a line of product was written.
• Aver herself, designed, voiced and built as a conversational avatar who learns a user’s baseline over time and explains what their health markers actually mean for them.
• A retrieval system ingesting thousands of medical documents, so answers are grounded in clinical literature rather than generated from a general model.
• Live integrations across multiple wearable and native biometric devices, normalised into a single health picture.
• Human-in-the-loop escalation, routing to medical professionals rather than replacing them.

Compliance built in, not bolted on

We designed and implemented the control environment for SOC 2 and built HIPAA safeguards into the architecture from the first sprint, not retrofitted once the product worked. We then guided the platform through readiness and into audit.

Doing it in that order costs more at the start and far less later. Access rules, data handling, retention and escalation paths were decided alongside the product rather than discovered at the end.

It also meant the team never had to choose between shipping and passing.

The brief

Unify a fragmented market (tracking, training, tele-vet, commerce and community) into one connected system for pet health. An avatar-led app that turns a continuous stream of pet data into something an owner can actually understand and act on.

FitFur arrived with the idea and not much else. We designed the characters, built them, and built the app they live in. The founder has been open about it since: he had an idea and no idea where to start.

The Challenge

Pet health has no equivalent of the clinical baselines that exist for humans. Raw biometrics off a wearable are meaningless to an owner without a model that can say whether this is normal for this animal. We had to define the product and the data science at the same time, then make the output simple enough for a non-expert to act on.

And the people reading it range from a vet nurse to a nine-year-old. One voice was never going to cover that.

What we built

• A device integration layer ingesting activity, heart rate, sleep and location from wearable hardware in real time, normalised into one record per animal.
• A longevity scoring engine that analyses an individual pet’s biometrics and turns continuous data into a signal an owner can understand.
• A cast, not an assistant. Kat, Adam, Cleo and Dash, designed, voiced and built so an owner chooses who explains their pet’s health to them. No other app of this kind has shipped multiple avatars. This is the first.
• Avatar-led interpretation, behavioural training support and wellness guidance in plain conversational language.

Built to grow

The cast is the growth model. Adding an avatar is a content decision rather than a rebuild, so FitFur can reach a new audience (a breed community, a clinic, a new market) without touching the product underneath. Cleo and Dash already do that for younger users, who get a character they’ll actually talk to inside the same system.

The platform extends the same way: toward vets, insurers, and whatever wearable hardware comes next.

Why we built it

Client work proves we can deliver against someone else’s brief. Mayson is what we build when the brief is ours: no compromises, no committee, our own standard as the only constraint.

Avatars Global has shaped Mayson across product definition, character and brand, marketing, software engineering and compliance.

It’s also the clearest answer to a question we get asked a lot, what would you do with nobody else in the room. This is it.

What it is

A purpose-built financial intelligence rather than a general assistant pointed at markets, with research, live market data, news and portfolio tools in one conversational interface.

Mayson is the face of it. You ask her something the way you’d ask a person, and the research, the charts and the news arrive inside the conversation rather than as four tabs you reconcile yourself.

Designed, voiced and built in-house, like everything else here.

Guardrails first

Building a conversational system that touches someone’s money means the restraints matter more than the capability. We designed ours in from the start:

• Paper trading mode, so a user can run the entire system without a cent at risk.
• No autonomous execution. Trades run on the user’s own criteria, and the user approves every one before it fires.

Capability is the easy half. The restraint is what makes it usable by someone who isn’t already an expert.

Clarity and trust

• Standing disclosure that Mayson is a research tool which can be wrong, and that anything it surfaces should be independently verified.
• Third-party news labelled as third-party, aggregated, attributed, and explicitly not endorsed.

Both exist for the same reason. A financial avatar that sounds equally confident about everything is worse than no avatar at all, so Mayson is built to be explicit about what it is, where something came from, and where its knowledge stops.