AI implementation for business: how Valorem First uses tools, agents and bespoke software

Valorem First, based in Glasgow, puts AI to work in three ways. AI Tools rolls out Claude, ChatGPT and Microsoft Copilot with the training and guardrails to make them stick. AI Workforce builds AI agents that run high-volume processes. Bespoke AI builds custom software, with AI doing much of the build. We get inside a client’s processes first, so what we build changes the day job rather than sitting alongside it as another tool nobody opens. Most engagements start with a Rapid AI Assessment.

Why do most AI investments fail to deliver?

Most AI investment fails on adoption and governance, not on the technology. Licences on their own deliver close to nothing.

Most organisations have already bought something: a handful of Microsoft Copilot licences, a ChatGPT subscription, a pilot that impressed in the demo. Then usage quietly falls away, because the tool was added to the work rather than built into it.

So we start somewhere else. Before we recommend or build anything, we look at how your teams really spend their week: which processes hurt, where hours disappear, what gets copied from one system into another. What we build is designed around that, which is why it gets used. If you’re still at the groundwork stage, our guide to preparing your business for AI adoption is the place to start.

What is the AI value curve?

The AI value curve is our five-stage model of AI maturity: Experimenting, Assisting, Delivering, Scaling and Compounding. Placing your organisation on it tells you what to do next, and what has to be true before you move on.

1. Experimenting. People try AI on personal accounts, with no policy and no visibility. There’s no return yet and real risk, so governance comes first: a readiness assessment and an acceptable use policy.
2. Assisting. Proper licences and training that landed, with AI helping people do their existing work. This stage lives or dies on adoption. You’re ready to move on when usage is embedded and someone has named a painful, high-volume process worth automating.
3. Delivering. AI doing the work rather than helping someone do it, with one or two agents live on real processes. Governance becomes live work, because the AI now touches your systems or customers.
4. Scaling. Agents across several processes, monitored, maintained and owned by someone. Your cost base stops tracking your volume, and ongoing support matters as much as the build.
5. Compounding. AI is assumed in how work is designed, not bolted on afterwards. Your customers can feel the difference.

The jump that matters commercially is from two to three, when AI stops saving time and starts creating capacity.

How does Valorem First approach AI implementation for business?

We offer three services, matched to where you sit on the curve. AI Tools suits stages one and two. AI Workforce and Bespoke AI come in from stage three, once AI starts doing work rather than assisting with it.

AI Tools: rolling out Claude, ChatGPT and Microsoft Copilot properly

Claude, ChatGPT and Copilot, with the training and guardrails to make them stick. It’s the quickest way to start, and usually where we begin.

Who it’s for. Businesses with little or no AI capability today.

What it’s worth. Time back for people you’re already paying, without changing a single system. No integration and no long IT project. We estimate 3 to 5 hours per person per week for knowledge workers who use it properly.

The honest caveat. This only works if people actually adopt it. That’s why training and guardrails are part of the price rather than an upsell.

What it looks like in practice:

• Meetings. Recording and minutes that write themselves, with actions pulled out and sent on.
• Document work. Claude Cowork handling the drafting, formatting and rewriting that eats an afternoon: reports, specs, client correspondence, tender responses.
• Knowledge bases. Notion set up so the AI can actually read it, which means people get answers instead of a list of search results.
• Analysis and reporting. Claude Code turning spreadsheet exports into clean analysis, without anyone having to learn a business intelligence (BI) tool.
• Month end. Finance teams using Claude for reconciliations, variance commentary and first draft reporting. The month starts with a draft, not a blank page.

AI Workforce: AI agents that do the work

Not AI that helps someone do a task, but AI that does the task.

Who it’s for. Businesses with painful, high-volume processes.

What it’s worth. Work that gets done without a person doing it. That’s capacity rather than time saved, and it holds as volume grows. We estimate 60 to 80 percent of the handling time comes out of a bounded, repetitive process. The strongest version of the argument is simple: your team stops growing headcount in step with volume.

What it looks like in practice:

• Voice. Agents that hold real phone conversations and hand structured output back into your systems. Culture discovery interviews, HR, sales qualification, first line support. Ours is called George. Ring him on +44 141 673 5900.
• Admin. An agent watches a mailbox, reads each job request, pulls the missing context from other systems, and emails a fully populated job ready to approve.
• Drawings. AI reads the title block, then renames and files each drawing to your naming convention. Days of document control, handled in the background at any volume.
• Estimating. AI reads drawings, specs and technical documents and pulls out quantities and scope. Your estimators keep the commercial judgement and price more opportunities each week, which shows up as revenue rather than cost saved.
• Your own data. Agents on top of your databases, so leaders can query live company data in plain English. No dashboard, no BI queue, no three day wait.
• Your own assistant. An agent that knows your business, sitting in Teams or SharePoint. It answers from your policies, contracts, drawings and project files rather than the internet, and everyone gets the same answer.
• Technical knowledge for sales. We ingest your documentation, then use George to interview the engineers who hold everything that was never written down. Your sales team can answer technical questions in the meeting, with guardrails on what reaches a customer.

Bespoke AI: custom software built with AI

When nothing off the shelf fits, we build it. Holiday management, surveying and quoting tools, client-facing apps that show project progress.

Who it’s for. Anyone who has outgrown the workaround, or is stuck in Excel.

What it’s worth. The process nobody has built software for is usually your real competitive difference. The alternative is a spreadsheet, a workaround, or a system that fits sideways. Because AI does much of the build, you get a working version in weeks rather than months, at a fraction of the cost of traditional development. And you own the code.

AI readiness, training and governance

Readiness, training and governance decide whether AI spend lands, because most AI investment fails on adoption rather than the technology. Readiness tells you where to spend first; policy and governance are what let anything go live.

• AI readiness assessment. Where you are now across six dimensions, and what to do first. Usually the way in.
• Training. Getting people genuinely using it rather than nodding through a demo. Role specific, not a generic hour.
• Policy and acceptable use. What staff can put into which tool, what needs a human check, and who signs off. Short enough that people actually read it.
• Governance and risk. Data residency, retention, and what the AI is allowed to say to a customer. The questions your IT and legal teams will ask before anything goes live.
• Ongoing support. Someone to ask. Agents drift when processes change, and models get updated underneath you.

AI in practice: what we’re working on now

These are live projects, with client details anonymised.

AI tools across a whole business. Claude rolled out to every team, a knowledge base for tender management, automated meeting notes, and AI-assisted take-offs feeding the estimating process. The pilot is finished and it’s changing how people work day to day.

Live company data in plain English. A Model Context Protocol (MCP) server on a reporting database, so people query live data in plain English instead of waiting on a report. The same teams are using Claude Code and Claude Cowork to prototype and build internally, at a pace they couldn’t manage before.

An agent running a shared mailbox. An AI agent reads incoming job requests, gathers the missing context from other systems, and sends the responsible person a fully populated job to approve. Twenty minutes of copying becomes a thirty second review.

Document control at scale. Thousands of files audited for accuracy and consistency in the background, at a volume nobody would attempt by hand.

Rapid AI Assessments. Completed for some organisations and under way with others, setting out where each one stands and what to do first.

What’s next: voice agents for sales and HR

We’re talking to clients about voice agents for out-of-hours sales enquiries, so calls that currently go to voicemail get answered and qualified. And about voice for internal appraisals and interviews, where every employee needs to be spoken to properly rather than sent a survey nobody fills in.

How to get started with AI

Most of our work begins with a Rapid AI Assessment, part of our AI adoption service. It tells you where you sit on the curve, what to do first, and in what order, so the first pound you spend is the one that returns most.

If you’d like to talk it through, get in touch with our team. Or ring George on +44 141 673 5900 and hear for yourself what an AI agent on the phone sounds like.