Each month, Lucky Saint’s London sales team have to visit over 400 pubs that serve its alcohol-free beer on draft. Until recently, planning these visits entailed manually grouping venues and plotting routes, taking at least an hour per day for each salesperson. Since they started working with London-based workplace AI training company ivee , the team have been using AI to cluster venues by location and automate their route planning. Ivee says the change has returned roughly five hours per week to each salesperson ay Lucky Saint , allowing the team to spend more time with pub managers and visit 20% more pubs each week.

Whilst this may sound modest compared to the autonomous agents and wholesale job redesign often promised by the technology industry, it also illustrates the problem ivee is trying to solve. Businesses are buying increasingly capable AI systems, then leaving employees to work out where those systems fit into their jobs.

Ivee’s answer is an AI Learning Companion that works alongside existing tools, including ChatGPT, Claude, Gemini and Copilot. It can improve and score prompts, recommend a more suitable model, build a reusable prompt library and, eventually, identify repeatable tasks that could become automations. The beta is being tested with selected companies and users, with a wider public release planned for 23rd October 2026.

Although the prompt improver is the main product, it opens the door to whether ivee can become a training layer between workers, AI models and the systems in which work actually happens.

AI Adoption Is Outpacing Employee Capability

The workplace AI gap is becoming an increasingly discussed topic. A 2025 Skills England report examined upskilling barriers across ten growth sectors, reflecting how AI capability has become a workforce issue rather than one confined to technology teams. Google’s UK AI Works pilots found that only a few hours of training could help participants use AI twice as much, while estimating average administrative time savings of a whopping 122 hours a year.

The key issue is that simply granting access to tools does not lead to useful adoption of them. It can produce unused licences at one extreme and uncontrolled experimentation at the other, with employees repeating weak prompts, choosing unnecessarily expensive models or placing sensitive data into tools that have never passed procurement.

Amelia Miller, co-founder and CEO of ivee, describes this gap as adoption debt. “From day zero, the minute a company rolls out a very expensive enterprise AI license, whether it be to Copilot, ChatGPT, Claude, Gemini, they’re already on the clock when it comes to demonstrating the ROI from that AI investment,” she said. “It’s the difference between rolling out these subscriptions and actually teaching your employees how to use these tools.” The term reframes poor adoption as an organisational design problem; when a company pays for technology without providing time, governance or role-specific training, disappointing returns cannot simply be attributed to reluctant employees.

Workplace AI Training Is Moving Into The Workflow

Ivee was founded by sisters Amelia and Lydia Miller (co-founder and CTO) after watching their mother struggle to return to work without the AI skills that employers were increasingly expecting. The company says it has since trained more than 100,000 people across over 1,000 teams, convincing the founders that classroom sessions and self-paced courses were a poor match to a technology that changes every few weeks.

“Gone are the days where we all sat down and spent 50 hours learning Microsoft Excel to get a shiny certificate, and then we just knew how to use it for the next 20 years,” Amelia Miller said. “This concept of micro-upskilling or micro-learning, learning on the job as you go, is how we learn AI now.”

With ivee, users can highlight a prompt and activate the companion through a keyboard shortcut. The system returns a rewritten version, scores the original for qualities including context, specificity and format, and then explains what made the revision stronger. Users can then replace the original prompt without moving into another application. The aim is to insert thousands of small lessons into real tasks, when the explanation is immediately relevant.

Where a prompt optimiser can make a worker more productive for a moment, the larger idea is that a learning product has to leave the worker more capable afterwards. Ivee’s website says each rewrite is accompanied by an explanation intended to help users start writing stronger prompts themselves.

Better Prompts Are The Starting Point For Workflow Automation

The beta includes prompt improvement, model selection, company and personal settings, a shared prompt library and an ROI dashboard, and Ivee says it can also detect unsanctioned AI tools. Team plans are expected to cost £16 per user per month, while a free tier will cover prompt improvement and token optimisation.

The company claims that more efficient prompts can reduce token use by an average of seven times. Lydia Miller explained that the calculation combines fewer conversational turns, shorter outputs and the ability to move straightforward work away from an unnecessarily advanced model. “It’s not always exactly seven times,” she said. “That’s kind of an estimate that we’ve done based on the tests that we’ve run and the figures that have come off the back of that. Sometimes it’s much, much more, and sometimes it’s a little bit less.”

A potential larger commercial opportunity goes beyond prompt correction, with Ivee testing a feature intended to recognise repeated manual work and suggest automations using tools that the organisation already permits. This could turn a poorly expressed instruction into a better prompt, then a reusable task and eventually a redesigned workflow. Looking back at the example from Lucky Saint, the value did not come from producing a more eloquent answer, but from identifying a recurring piece of administration and changing how it was performed.

AI Literacy Still Requires Independent Thought

There is a clear tension around placing a corrective layer between a person and an AI model; if the software continually supplies context, structure and language, is the employee learning, or just becoming dependent on another tool? Lydia Miller says ivee has encountered a limit when users have not formed an idea of what they want. “We can’t read your mind and we don’t know your business objectives. We can’t make up the work for you,” she said.

When the system spots that there is some missing context, it can insert placeholders asking the user to provide their own position. “AI will generally pick the average response,” Miller added. “And if everyone’s using AI to pick the average response, the responses all sound the exact same.” AI literacy cannot just be reduced to a series of elaborate prompts, instead it requires knowing when a model lacks context, when an output is generic and where professional judgement remains necessary.

Ivee also allows organisations to establish shared style rules, while individuals can add their own preferences. Miller calls these its “anti-AI-slop rules”, joking that “if anything of ours ever went out with the word ‘quietly’ in it, I think we’d have to have a serious team meeting.”

AI Training And Workplace Monitoring Share The Same Data

Ivee’s ability to coach employees depends on the ability to observe aspects of how they use AI. This same information can help managers identify weak adoption, escalating token costs, security breaches and shadow AI, while also creating a sensitive boundary between enablement and workplace monitoring. Crucially, the company says sensitive information can be redacted locally before it reaches either ivee or the underlying AI model.

The companion only accesses a conversation when a user activates it, according to Lydia Miller, and stores the selected prompt rather than the full conversation. The surrounding chat can be processed to determine whether a task is repetitive or the user should start a new thread. For employers, she says that the dashboard presents anonymised, aggregated patterns rather than individual prompt text. Her example is deliberately pointed: “They could, if they wanted, improve a prompt saying, ‘Write me a resignation letter from this company,’ and their employer will not be able to see that.”

This boundary is a vital one to maintain as the product grows in capability, as companies will want evidence that their investment is delivering a return, whilst workers will want to know what is processed, stored and surfaced to management, with ivee’s usefulness hinging on earning trust from both sides.

Can Embedded Workplace AI Training Close The Adoption Gap?

Ivee is initially targeting small and medium-sized businesses that have purchased AI tools but lack dedicated adoption teams or extensive training budgets. These companies are looking for practical productivity gains, but may have fewer resources to create policies, evaluate models and redesign jobs responsibly. The founders frame AI fluency as a form of career resilience: “What we don’t want is people sleepwalking into unemployment because they don’t have the skill set required because they were at a company that was falling behind,” Lydia Miller said. “We’re really trying to help that 99% become AI-native, so that they can actually use the tools. They’re not daunted by them, and they are therefore very valuable employees.”

However, learning to use AI may not make every worker “unfireable”, and businesses still have to decide how the time released by automation is distributed. The Lucky Saint example works because saved time can be redirected towards relationships and revenue, but in other roles, the outcome could be higher workloads or fewer people.

The ambition for ivee is a big one, as summarised by Amelia Miller, “What Lydia and I have leant on is our backgrounds, mine in psychology, and Lydia’s in investing in AI companies, to figure out how can we actually crack helping an individual employee learn AI at the rate that they need to, in a very personalised way, inside the tools that they’re already using, rather than having to rely on a really fancy corporate training partner or really expensive certification program, because you just can’t do it at the pace that AI is moving at the moment.”

The key question now, beyond whether ivee’s workplace AI training tools can improve an individual prompt, is if they can turn millions of small interventions inside the working day into lasting capability for workers and measurable learning for organisations.