The highest AI adoption of any sector. Engineers calling models directly.
Technology firms lead UK adoption, engineers call model interfaces straight from code, and customers now ask how you control AI in your product and in your own work.
of UK information and communication firms use AI, the highest of any sector
ONS, 2026of AI cost goes to premium models, up from 6% in ten months
Ramp, 2026, vendor dataset of its own customersof UK businesses using or considering AI have cyber practices that address AI risk
DSIT Cyber Security Breaches Survey 2025/26, as reported by secondary sourcesYou are responsible for the AI you use and the AI you ship.
The Information Commissioner's Office (ICO) shared preliminary views on agentic AI in February 2026: organisations remain responsible for what these systems do with personal data. New rules on automated decisions under the Data (Use and Access) Act 2025 took effect on 5 February 2026, and the ICO is preparing a statutory code of practice on AI. For the EU, the Council and Parliament reached a provisional deal on 7 May 2026 to delay high-risk AI Act duties to 2 December 2027 (stand-alone systems) and 2 August 2028 (AI inside regulated products); formal adoption should be confirmed.
Summarised from ICO publications and law-firm reports on the EU deal. Not legal advice.
Six problems that keep coming back.
Drawn from public surveys and what technology buyers now ask suppliers.
Engineers call model interfaces from code, and the bill arrives after the month ends. Premium models take most of the cost even when a cheaper one would do the job.
Coding assistants and chat tools receive proprietary code, keys and customer data. Few teams can show what left the building.
If your product sends customer data to a model, your customers' security teams will ask which model, where, and who can see it. Answers scattered across code are hard to give.
Enterprise buyers ask about AI use, human oversight and incident handling. Teams answer from memory, slowly, and sometimes inconsistently.
A provider retires or updates a model and outputs shift. Without a test and a change record, support and engineering find out from customers.
AI agents act with credentials. Scoped access, dedicated identities, logs and a stop switch are now basic expectations, and often missing.
The cost shows up in margin, in deals and in trust.
Unmetered AI spend erodes gross margin as usage grows. A questionnaire you cannot answer delays a contract by weeks. A leak of customer data through an AI feature costs the customer, the contract and the reference. Each is small at first and expensive when it compounds.
Control is a feature customers will buy.
Teams treat governance as a brake on shipping. In practice one set of rules for model calls, with budgets, redaction and a record, is one change for engineers and gives sales a ready answer to security reviews. The same evidence serves procurement, your own board and any later certification you choose to pursue.
- Code generation and review
- AI features inside your own product
- Support and success automation
- Sales research and outbound
- API spend with no budget line
- Customer data in product AI with no record
- Procurement blocked on evidence you do not have
GenWA applies your rules to model calls and meters them with per-team budgets, records product AI in the log, supports sandbox testing before model switches, and prepares mapped evidence for customer reviews. Controls apply to connected, routed tools.
Four agents, trained for this sector.
The rules, as controls.
- ISO/IEC 27001 and 42001 (readiness)
- SOC 2 trust criteria
- UK GDPR and the Data (Use and Access) Act 2025
- Consumer Rights Act 2015
- Online Safety Act (where applicable)
- EU AI Act (where you serve the EU)
Built around product teams, platform teams and founders.
Technology firms range from founder-led start-ups to scale-ups with security teams. We shape each route to that.
An independent audit including direct model use and spend, with first findings in 48 hours and a plan for your leadership team.
Start the health checkAdvice: policy and customer assuranceAn AI policy and answer bank for security questionnaires, written with your security and legal leads.
See advisoryProduct: rules and evidenceRules, Log, Evidence Packs and Compliance Mapping, hosted in your AWS or Azure account, or GenWA AWS as agreed.
Find the right productProduct: agents at workSupport automation, growth operations and invoice processing with named owners and spend caps.
Meet the teamIncubator: one function, twelve weeksWe define, build and run one controlled function, such as support triage. You own the intellectual property, prompts and record.
See the incubatorAccelerator: for AI-first foundersBy application only, ten places a year. Senior AI leadership time, advisory and platform access, paid for by equity, project, retainer or licence, with controls from day one.
See the acceleratorWhat Guard and Evidence typically hold for this sector.
Control-mapped, not certified. We prepare the evidence; accredited bodies issue certification.
The international standard for managing AI. Controls mapped to its clauses for readiness.
The information security standard. AI-specific controls mapped alongside your existing set.
The US-origin service-controls report many enterprise buyers request. Controls mapped to its criteria.
Data protection and automated-decision safeguards, with redaction and review evidence.
Where you place AI on the EU market.
The UK government-backed baseline cyber scheme that customers often require of suppliers.
Three role holders, three different problems.
Owns engineering and cannot see model use. GenWA gives one set of rules with budgets, model choice and a record, without slowing engineers down.
Answers customer reviews and the board. GenWA keeps the register, rules and evidence pack that turn a six-week questionnaire into a document you can send.
Sees AI cost as a growing line with no owner. GenWA shows spend by team, task and model, with caps that pause work rather than bill past a limit.
From this sector.
What does GenWA do for a technology company that ships and uses AI?
Rules and Cost Meter that apply your policy to model spend with team budgets, a log for product AI, sandbox testing before model changes and mapped evidence for customer security reviews. It also helps keep source code and secrets out of unapproved tools on connected routes. We promise no fixed saving.
Will our engineers resist new controls?
It is designed to be one endpoint change, and it gives engineers model choice, budgets and a record. We have no client adoption evidence to quote yet, so we suggest judging it in a pilot with one team.
We already use a cloud model platform. Is GenWA redundant?
No. GenWA works alongside the platform you chose and adds rules, redaction, budgets and a log across approved providers and private models.
How do we answer customer AI security questions?
Compliance Mapping and Evidence Packs map your controls to ISO/IEC 42001 readiness and other frameworks buyers cite. We prepare the evidence; accredited bodies certify. Check which standards your customers actually ask for.
How do EU AI Act rules affect a UK technology company?
It depends on whether you place AI systems on the EU market or serve EU users, and on the risk category. A provisional EU deal of 7 May 2026 would delay some duties, but final adoption should be checked at source with a legal adviser.
How do we keep source code and secrets out of unapproved tools?
Connect your tools to GenWA with redaction and rules on what may be shared, and use the register to see what else is in use. The log records requests. Coverage depends on the integrations scoped in the audit.
How does a SaaS company start?
Take the AI health check, then pilot Guard on one team. First findings arrive within 48 hours and the audit with a 90-day plan within ten working days. The audit fee is credited in full against year one of a platform plan.
Where does your AI spend go, and what can you tell your customers?
First findings within 48 hours of the health check. Audit fee credited in full against year one of a platform plan.