Real AI projects, not buzzwords
How ABCOM is using AI today, for clients and in our own work, to save hours, keep data private and turn conversations into clear plans.
- Private where it matters
- A human decides
- Consent first
Game-changing AI
2 hrs → 15 min meeting follow-up time
AI we have delivered
Four real examples: one client project, two ways we use AI in our own business, and one community STEM project.
Faster data access and private AI for a London accountancy practice
The challenge: the multi-site practice held its data in a hosted environment. Staff experienced slow sessions, lag and frequent disconnections, and the firm needed secure, auditable access for external consultants. They came to us after hearing about our work with another accountancy practice in Sussex.
What we delivered: a dedicated high-spec server, running a virtual machine with an offline large language model for report crunching, so substantial reporting material is analysed inside the firm’s own environment and never sent to a public AI service. We also improved cross-site and consultant access, hardened the firewall and added endpoint detection and response (EDR) and identity threat detection and response (ITDR).
Early results: systems are faster, the frequent disconnections have stopped, and the firm has a private platform for developing its reporting workflows.
- Private, offline AI
- Faster, stable access across sites
- EDR and ITDR protection
From a 60-minute meeting to a shareable plan in minutes
The situation: a 60-minute discovery conversation with the partners of a local dental surgery. With their consent, we recorded the whole meeting using an AI voice recorder.
What AI produced: back at the office we generated clear action points and follow-ups, a mind map to show the discussion visually, and a RAG knowledge base of the conversation. That is a searchable, question-and-answer version of the meeting, shared with the client so they can check exactly what was agreed.
The result: preparation that would normally take around two hours took 5–15 minutes, and both sides had the same accurate record.
- Recorded with consent
- Action points and follow-ups
- Mind map and shareable RAG
60-minute discovery conversationRecorded with consent, using an AI voice recorder
What came out of it
- Action points and follow-ups
- A mind map of the discussion
- A searchable, question-and-answer version of the meeting
From a backlog of website fixes to done in days
The situation: like many small businesses, we had a long list of website improvements that never reached the top of the to-do list.
How AI helped: we used AI under close supervision, with clear goals set by Yann and every decision made by a person.
The result: more than 80 improvements went live in about two days, work that would normally take weeks around running the business. Every change was reviewed, approved and backed up before it went live.
- 80+ changes in about two days
- Every change human-approved
- Weeks of time saved
80+changes in about two days
Work that would normally take weeks around running the business
- Reviewed
- Approved
- Backed up
- Live
AI-assisted sound design for a school plant-music project
The situation: as a STEM Ambassador for Mid Sussex Science Week, we helped Bolney Primary School turn signals from living plants into music. A Scion sensor turned tiny electrical changes in a fern, cactus and rubber plant into MIDI, which drove instruments in Ableton Live, while pupils added birds, water, weather and drums from a Novation Launchpad.
How AI helped: we used AI to help capture and create the sounds loaded into Ableton, the birds, water, weather and ambient textures, so we could build a rich, classroom-friendly sound palette quickly.
The result: pupils played a live “plant orchestra”, and ABCOM and the BHBPA plan to donate a Scion to the school so they can keep experimenting. Read the full story.
- Community and STEM
- AI-assisted sound design
- Human-led, child-friendly
Tap a plant to hear a note
- A living plant
- Scion sensor
- MIDI
- Ableton Live
How we use AI safely
The same principles we apply for clients in our AI Readiness Sprint.
Consent first
We only record or process people’s information with their agreement.
Private where it matters
Sensitive data stays in your own environment, such as an offline model, rather than a public AI tool.
A human decides
AI drafts and analyses; people review, approve and stay accountable.
Governed data
Permissions, sensitivity labels and Microsoft Purview controls before tools like Copilot are switched on.
Thinking about AI for your organisation?
Start with a clear view of your data, security and permissions. Our AI Readiness Sprint gives you a practical plan before you roll out tools such as Microsoft 365 Copilot.