acAIberry

Research & Development

Everything we sell with confidence today, we were once unsure about.

R&D at Acaiberry is not a marketing department with a lab coat.

It is a standing budget, a small team and a habit of building the thing before a client asks for it, so that when they do ask we can answer from experience rather than from a vendor's slide deck.

The output is a running view of which technologies are ready for production work, which are worth piloting, and which we are only watching. That view is below, and it is the honest version.

See what we are tracking
State of play

13

tracked

4

in production

4

positions

Q3

July 2026 review

Reviewed quarterly. Technologies move right only when we have evidence, and things move left as well.

Innovation as a core capability

Somebody has to absorb the risk of finding out

We pay to find out first

A client asking whether a new technology will work is really asking someone to spend money finding out. We would rather carry that cost ourselves, on our own time, and arrive with an answer we have already paid for.

The board represents real cost

Each position cost something to establish: a prototype that worked, a prototype that did not, or a pilot that ran for six months before we understood what the data was telling us.

Shorter work, fewer surprises

When a request lands in an area we have researched, we can size it in days rather than weeks, name the risks specifically, and start from working code rather than a blank repository.

Emerging technology research

What we think is ready, and what is not

Four positions: watching, experimenting, piloting, in production. A technology only moves right when we have evidence, and things move left as well.

Step 01

Watching

No client work yet. We are tracking the signal, standards and economics, but we are not selling this as a client solution until the production case is clearer.

Quantum computing

No near-term production case for the organisations we serve, and we will say that plainly rather than sell a workshop about it. We track it for one practical reason: the timeline it sets for cryptographic migration. If your organisation holds records that must stay confidential for a decade or more, that timeline is already relevant. What would move it: A credible cost advantage on an optimisation problem a client actually has.

Applied research & technology transfer

Five stages, and most ideas stop early

Research earns its keep by reaching production, but the route is deliberately narrow. Each stage has an exit condition, and failing one is a cheap outcome rather than a bad one.

Feasibility studies

Whether the thing can work at all, given your data, your constraints and what the field can currently do. Delivered as a written answer, including no.

Proof of concept

The riskiest assumption tested in isolation, on real data, in weeks. Built to be thrown away once it has answered its question.

Prototype development

A working slice that people can use and react to, which surfaces requirements no workshop ever produces.

Technology validation

Independent evaluation of a platform, model or vendor claim against your own workload, benchmarked on cost and quality rather than a datasheet.

Pilot programmes

Limited live deployment with success criteria agreed in advance, run long enough to see seasonal and operational reality.

Transfer to production

Prototype code rewritten to production standards, monitoring added, handover documented. The stage most innovation programmes never budget for.

Feasibility studies

Whether the thing can work at all, given your data, your constraints and what the field can currently do. Delivered as a written answer, including no.

Proof of concept

The riskiest assumption tested in isolation, on real data, in weeks. Built to be thrown away once it has answered its question.

Prototype development

A working slice that people can use and react to, which surfaces requirements no workshop ever produces.

Technology validation

Independent evaluation of a platform, model or vendor claim against your own workload, benchmarked on cost and quality rather than a datasheet.

Pilot programmes

Limited live deployment with success criteria agreed in advance, run long enough to see seasonal and operational reality.

Transfer to production

Prototype code rewritten to production standards, monitoring added, handover documented. The stage most innovation programmes never budget for.

Product innovation

We build our own products too

Building and running products for ourselves teaches things client work does not: what breaks at scale, what support actually costs, and which architectural shortcuts you regret.

01

SaaS platforms

Multi-tenant products taken from first commit to paying users, including the billing, onboarding and support machinery nobody demos.

02

AI products

Applications where the model is the product, built with the evaluation and cost discipline that keeps them viable past the first thousand users.

03

Enterprise platforms

Internal systems built once and reused across clients, which is how a two-year data platform becomes a four-month one.

04

Digital services

Public and citizen-facing services designed for low bandwidth, older devices and users who will not read instructions.

Collaborative research

Research we could not fund alone

The questions worth answering are usually bigger than one company's budget. We work inside consortia, grant programmes and university partnerships where the cost and the findings are shared.

Universities

Joint supervision, specialist equipment access, and students who ask the questions we have stopped asking.

Government

Public sector pilots, digital service prototypes and evidence for policy that needs a technical basis.

NGOs

Field research in low-connectivity settings, where the constraints are harder than anything in a lab.

Research institutes

Domain expertise in health, agriculture and environment that we do not pretend to have in-house.

Industry partners

Access to real operating environments, real failure data and the problems worth solving.

International organisations

Multi-country programmes where a solution has to work in several regulatory and language contexts at once.

Innovation lab engagements

An embedded team running your experiment pipeline for a fixed period, so the habit stays inside your organisation.

How the work gets funded

Research grants, innovation challenges, co-funded pilots, public sector innovation routes and technology demonstrations. We are open to consortium invitations where the research question needs engineering to answer it.

Knowledge leadership

What we learn, we publish

Research kept private is a marketing asset. Research published is an argument other people can check, which is a higher standard and a better reason to trust us.

Signal 06White papersSignal 05Technical publicationsSignal 04Conference presentationsSignal 03Industry workshopsSignal 02Executive briefingsSignal 01Training programmes
Signal 01

Training programmes

Structured upskilling for client engineering and analytics teams, because handover only works if someone can maintain what we built.

Engineering teams
Analytics teams
Maintenance skills
Handover support
Operating playbooks
Capability transfer

Innovation methodology

How an idea moves through the lab

Seven steps, each with a decision at the end of it. The point of the structure is not to guarantee success. It is to make abandoning something cheap and unembarrassing.

STEP
01

Opportunity identification

A recurring client problem, a technology shift, or a question we keep failing to answer well. Candidates are written up in a paragraph, not a deck.

CLIENT SIGNALS • TECHNOLOGY SHIFTS

01/07

WHY THIS MATTERS TO YOU

You get the benefit of experiments you did not pay for

Emerging technology at reduced risk
Faster experimentation from working code
A path from concept to production
An honest read on hype
Access to a wider research network
Capability that stays with your team

Open questions

What we are working on right now

We keep finding the accuracy cliff in a different place for every client's lighting and defect mix, which makes hardware selection a guess. We want a cheap test that predicts it before anyone buys a board.

Bring us a question we cannot answer yet

An executive briefing is ninety minutes and comes with no proposal attached. A feasibility study is a few weeks and ends in a written answer either way. If you are assembling a consortium or writing a grant proposal that needs a technical partner, that conversation is worth starting early.