Coeus · AI insights platform · South Africa

Idea to beta in six months.

Coeus set out to give small businesses the market intelligence only large ones can afford. What existed was a thesis and a founding team - the MVP had to unlock funding, prove demand, and be worth building on afterwards.

A live product, real users, and a price that covers what it costs.

A Coeus insights report on a laptop: a short-form winter food and beverage trends report prepared for a named cafe, drawn from twelve sources over a two-week window.

Client: Coeus
Sector: Technology · AI market intelligence
Role: Strategy, product and build
Stage: Concept to funded beta
Expertise: Product Strategy, Product Design, Full Stack Development, AI and Data Engineering, Interface Design, Cloud and DevOps Engineering

The need

Research priced for the enterprise. Needed by everyone else.

Coeus is an AI startup with a single conviction: small and mid-sized businesses decide in the dark, and not because the information does not exist. It exists. It is simply priced, packaged and written for companies a hundred times their size.

The founders had the thesis and the domain knowledge. What they needed was a partner to turn it into something buildable and then build it - an MVP market-ready enough to unlock funding and prove demand, without being the kind of prototype you throw away the moment it works.

Three things, all true at the same time

Fast enough to matter
Six months, not eighteen. A research product that lands after the funding window has closed is a different product, in a different market, with a different set of competitors already in it.
Priced for a cafe, not a corporation
The whole thesis is affordability. If the monthly number makes an owner-operator flinch, the idea has failed on contact, no matter how good the output is.
Cheap enough to serve
Every generated report spends real inference. A product that is affordable to buy and expensive to run does not have a pricing problem, it has an arithmetic problem - and arithmetic does not improve with volume.

The Coeus onboarding screen welcoming a new user to the market information tool for SMEs.

The Coeus plan picker, showing a free tier alongside paid tiers and explaining that generating a report spends report tokens.

What we delivered

An insights engine, not a chat window.

Coeus reads the web and social for your industry twice a day, files what matters into the topics you have told it you care about, and turns any of those into a short-form research report addressed to your business by name. We took it from an empty repository to that - and treated the product, the pricing and the infrastructure as one decision rather than three, because on an AI product they are.

Insights Hub - twice-daily roundups of what is moving in your industry
Pursuits - standing topics the engine curates against on its own
Coeus Assist - ask in plain language, answered from your own filtered insights
Reports - short-form research, sourced, dated and addressed to the business that asked
Token metering, so what a customer pays tracks what their usage costs to serve
Analytics wired in from the first build, so beta produced evidence and not opinions

Coeus Assist answering a question about festive-season buying in Cape Town, with the option to turn the answer into a Pursuit, beside the Food and Beverage insights hub.

Ingest
Web and social, twice a day
Insights Hub
Filtered to your industry
Pursuits
The topics you track, curated automatically
Reports
Short-form research, on demand

Each stage narrows the one before it. By the time a report is generated the model is reading a curated set rather than the open web, which is what makes the output specific enough to be useful and the cost per report predictable enough to price.

How we built it

Six months is a scoping decision, not a sprint.

Nothing here got built faster by working harder. The date was met by deciding early, and in writing, what the MVP would not do - and by settling the questions that get expensive later, like what a report costs to produce, while they were still cheap to answer.

01
A roadmap that says no
Scope was cut against one test: does this help prove demand? What failed it was written down and deferred rather than re-argued every sprint, which is where roadmaps usually leak their time.
02
Pricing designed with the product
The tiers, the free entry point and the token mechanic were settled while the architecture was still moving, so what was being built and what was being sold could not drift apart.
03
Infrastructure tuned to cost per report
The AI stack was built to carry large data loads at a running cost the business model could absorb. On a product that spends money every time someone uses it, that is not an optimisation - it is the margin.
04
Beta as an instrument
Structured testing with real users, and tracking in place from the start, so the next phase gets chosen on what people did rather than on what they said in a call.

The My Pursuits screen: a grid of tracked topics - Cup Culture, Bowl Goals, Farm-to-Fork Trends, Festive Feast Trends - each showing how many articles the engine has curated into it.

The Coeus insights hub for the food and beverage industry, with trending web articles beside twice-daily roundups and a menu for filing an insight into a Pursuit.

The weekly digest on a tablet, listing newly curated articles by Pursuit with a link through to each one.

What beta proved

One number, and the two things it bought.

0 monthsfrom first conversation to a beta in real users' hands

Validated with real users
Structured beta testing put the assumptions in front of people with no reason to be kind about them, and the feedback set the next phase rather than decorating it.
Built to scale with demand
The infrastructure takes on load without the running cost climbing in step, which is the difference between a demo that impresses and a business that survives being used.

The MVP is live and in use. What beta proved is not that the idea works in principle - it is that it works at a price a small business will pay and at a cost the company can carry.

The word from the client

“I can confidently say I've never met a more diligent, world-class product team - their work ethic and methodology are second-to-none. They're laser-focused on building products users will love and add immeasurable value to the creative process, from start to finish.”