AI isn’t waiting for our job descriptions to catch up. Dr Andy Packham

Seven in ten employees say they use AI at work, but only about four in ten feel satisfied with the training they’ve received. Meanwhile, pilots and proofs-of-concept (PoCs) are everywhere, yet an estimated 80%–90% of those AI pilots never make it to full production. One big reason is what I call the “pilot paralysis” problem: many pilots focus on proving a technology works, but don’t prove the full concept. The “concept” in proof-of-concept means the whole idea – the process, workflow, people and skills needed to succeed at scale. If a PoC doesn’t validate those elements, it’s not truly proving the concept at all – it’s just a tech demo. No wonder so many promising pilots stall before they ever deliver real value.

Yet that gap between potential and outcome can be closed—I’ve seen it happen. At HCLTech, we’ve helped clients turn AI experiments into tangible results. For example, in a legal due diligence process, AI-driven automation cut the review time by half and in a healthcare project, initial clinical document reviews were reduced by 88%. These aren’t theoretical wins; they’re real-world outcomes – the kind you can take to your CEO. And they happened because we paired the technology with the right skills, leadership and processes to operationalise it. In other words, we proved the concept, not just the tech.

Crucially, bridging this pilot-to-production gap isn’t about technology alone — it requires executive sponsorship and a culture shift. Recently, I discussed this challenge on a podcast with two experts on the front lines of AI skilling: Justin Slade, Microsoft’s Director of Business Strategy for Partner Skilling and Vinutha Rao, Senior Vice President at HCLTech, leading enterprise delivery and talent transformation. Our conversation reinforced a clear theme: getting from AI pilot to AI productivity comes down to people and skills. Technology may spark the fire, but it’s skilled people (with leadership backing) who stoke it into a sustainable blaze. When leaders champion AI learning and align it to business strategy, it sets the stage for cultural change and accelerates the whole organisation. In this blog, I’ll share key insights from that discussion — including six action steps — for executives eager to see their PoCs translate into production outcomes.

Why AI PoCs stall

Let’s unpack why so many AI initiatives get stuck in pilot paralysis:

Velocity shock. The pace of change in AI is dizzying. As Vinutha noted, “IT today is moving faster than one can imagine,” and what worked yesterday is not what’s going to work tomorrow. This rapid evolution can outpace an organisation’s skillset. If your workforce can’t keep up, even the best pilots will struggle to reach production.

Legacy skills and culture. Organisations often focus on upgrading technology but overlook the need to upgrade skills and mindset. Dealing with legacy infrastructure is important – but dealing with legacy skills and culture is just as vital. If teams are still operating with yesterday’s assumptions, workflows and comfort zones, even the most advanced AI tools won’t gain traction. Overcoming this requires visible leadership support and a shift to a culture of continuous learning. Teams need to feel that embracing new skills is expected and valued from the top.

Generic learning. Too often, training programmes are broad and theoretical, decoupled from business impact. We used to count certificates or hours of training and assume we were making progress. But mere exposure to AI concepts isn’t enough. If learning remains generic, employees might know about the technology but not how to apply it to solve real problems. Both Justin and Vinutha emphasised moving away from one-size-fits-all upskilling. Skilling should be outcome-driven and role-specific, not an academic exercise. When training is too generic, pilots falter because teams aren’t prepared to adapt the tech to their domain or workflow.

Pilot paralysis. This is the trap where a pilot proves to be a tool in the lab but fails to transition into the business. Often, it’s because the project was treated as a tech demo in isolation. It might have shown a model’s accuracy or a chatbot’s capability, but it didn’t address whether the organisation is ready to integrate it into real processes. In other words, it proved the tech, not the concept. Without an executive sponsor, a clear business owner and a plan for process change, the PoC has no bridge to production. It just lingers in limbo. We need to plan pilots from the outset as if success is expected — with end-users, business metrics and operational considerations in focus.

Governance lag. Enterprises rightly worry about governance – security, compliance, ethics. But governance frameworks often lag behind the pace of innovation. By the time risk and compliance teams figure out how to govern that clever new AI pilot, the technology (or the market need) may have shifted again. This misalignment can either slow projects to a crawl or, if ignored, lead to risky deployments. The solution is to bring governance and ethics experts into the conversation early. This ensures that responsibilities and guidelines are clear from day one, preventing last-minute roadblocks and building trust in the outcomes.

Domain–decision gap. AI solutions don’t live in isolation; they serve business domains. A common challenge is the gap between technical teams building AI and the domain experts who use the insights. If your AI team develops a great predictive model for, say, supply chain, but the supply chain managers don’t trust or understand it, the project will falter. Conversely, if business experts aren’t AI-literate, they might not even know what to ask for. Bridging this gap means training your tech people in the domain context and upskilling your domain professionals in AI basics. Vinutha described it as providing “adjacency skills” – adding business context to tech roles and vice versa – so that decision-making power moves down to where the expertise lies. When domain experts and AI specialists work in tandem from the start, PoCs are far more likely to tackle the right problems in a usable way.

These challenges can seem daunting, but they all share a common root: a skills gap and a culture gap. And that’s something we can tackle. Below are five action steps to accelerate skilling in your organisation and create the culture needed to turn AI pilots into scalable production successes.

Six action steps to accelerate AI skilling

If there’s one thing I took away from our conversation, it’s that AI skilling must be treated as a strategic business initiative driven by leadership, not a checkbox exercise. The following six steps can help executives supercharge their AI skilling efforts and, by extension, their AI outcomes.

1. Training must cover everyone in the value stream. One reason pilots fail is that training is limited to the tech team, leaving out sales, operations, finance, HR and other business functions. For AI to deliver real value, everyone involved in the process – from engineers to end-users – needs to be included in the skilling journey. This is where executive sponsorship is key: leadership should mandate and model that AI learning is company-wide. In practice, think beyond IT. Sales teams need to know how AI can support deal qualification and customer insights; marketing teams should learn to use AI for campaign targeting and personalisation; finance teams, for forecasting and anomaly detection; HR teams, for talent analytics and workforce planning; and so on. When skilling is inclusive and cross-functional, adoption improves across the board. It sends a cultural signal that AI isn’t just an “IT project” — it’s everyone’s business. The goal is to build confidence and capability at every level, so AI becomes a trusted tool in each department’s toolkit, not a mysterious project in the corner.

2. Tie skilling to business KPIs. Don’t train for training’s sake – train for specific outcomes. Every skilling initiative should connect to a business KPI that top leadership cares about. For instance, if your CEO is focused on customer satisfaction, link your AI training programme to that metric (e.g., “learn to use AI tools to cut customer response times by 30%”). By anchoring skilling to key performance indicators, you ensure it serves tangible business goals. Justin described skilling as a “strategic growth lever.” That means success is measured not by how many people got certified, but by metrics like faster sales cycles, higher win rates, lower costs or improved Net Promoter Scores. This approach demands executive involvement: when leadership drives the learning agenda and ties it to strategy, it sets the stage for real cultural change.

3. Go role-based, not generic. Different roles use AI differently. The most effective skilling is tailored to job roles and day-to-day tasks. This is where culture meets practicality: it shows respect for employees’ time and needs. Train your sales reps on using AI for lead scoring and proposal drafting; train your customer support agents on AI-assisted service tools; train your project managers on AI for risk forecasting; train your software engineers on AI-powered development tools; and train your department heads on interpreting AI-driven analytics to make decisions. Role-based learning paths ensure each person learns in context, making it immediately relevant. When someone can apply a new skill the next day, they retain it and feel the impact. In my experience, this tailored approach builds confidence and drives adoption – people see AI not as an abstract concept but as a concrete aid in their job. An added benefit: it helps identify gaps. If a certain role struggles even after training, you may discover a workflow issue or a need for more support, which you can address before scaling up. In short, personalise the journey – it’s more engaging and far more effective than any one-size-fits-all programme.

AI itself can play a powerful role in enabling this. Intelligent learning platforms can analyse job roles, performance data and skill gaps to generate tailored skilling plans for individuals and teams. This means employees aren’t just handed a generic course list — they’re guided towards the most relevant content, at the right depth and in the right format. It’s a smarter, more scalable way to personalise learning across the organisation.

4. Train in real business environments (on real work). The most powerful learning is by doing – especially by doing real work. I’m a big believer in moving skilling out of the classroom and into the field. This means integrating learning with actual projects and use cases, rather than relying on slide decks or toy examples. In practice, run hackathons and labs tied to live business challenges. When people train on real data, real problems and with real stakes, two things happen: you often create something useful (even as a proof of concept) and the team gains practical experience that sticks far better than any abstract tutorial. Just as importantly, train together as cross-functional teams. Everyone involved in the transformation–technical and non-technical – should share the learning experience. When a data engineer, a marketing analyst and a compliance officer sit in the same AI workshop, they build shared understanding and trust. They learn each other’s concerns and ideas and they’re more likely to work seamlessly when that AI solution goes live. AI projects don’t succeed in silos; they succeed when the full value stream is engaged. By learning in a collaborative, real-world setting, you not only develop skills, but you also cultivate the teamwork and buy-in needed to carry pilots into production.

5. Build internal champions and peer learning communities. In every organisation, you’ll find AI enthusiasts and early adopters – these are your champions. But to truly harness their energy, think beyond a few “AI heroes” doing side projects. Foster peer learning and coaching networks. Encourage your champions to form communities of practice (an internal “AI Guild,” for example) where they can experiment, share successes and coach their colleagues. These peer-driven communities can be a game-changer: they amplify early wins, reinforce skills and turn individual know-how into collective progress. Champions in a vacuum can burn out or go unnoticed; champions connected through a community can spark a movement. Give them air cover and a platform – perhaps regular show-and-tell sessions or an internal Teams channel to swap tips and answer questions. Crucially, allow these communities to be self-guiding. Often, the people on the frontlines will spot the next skill gap or AI opportunity before management does. By heeding those insights, you ensure skilling stays agile and relevant. In short, nurture an environment where employees learn with and from each other. It’s the difference between a workforce that’s waiting to be taught and one that’s hungry to explore. And from what I’ve seen, the latter is a hallmark of an AI-ready culture.

6. Teach how to make AI work for you – not how AI works. This is perhaps my favourite mantra to instil. We must shift the training focus from the technology itself to the application of that technology in our jobs. Most employees don’t need to dive into the maths of neural networks; they need to know how to leverage an AI tool to get better results. As I said on the podcast:

It’s not only about figuring out how AI works but also figuring out how I can use AI to work. Dr Andy Packham

All these steps reinforce each other. Cover everyone and you naturally encourage peer learning. Tie learning to outcomes and you automatically go practical and role-specific. Lead from the top and communities will flourish at the bottom. Together, they create momentum – a continuous upskilling engine that primes your organisation for AI success. I often call this building a skills flywheel.

Building a skills flywheel

One-off training won’t keep pace with AI’s evolution. Frontier Firms understand that AI skilling is not a one-shot deal but a continuous investment. I like to think of it as a skills flywheel: an engine of constant learning and application that builds momentum over time. How do you create it? By operating in short, iterative cycles – say, 90-day loops. Every quarter, identify a few priority AI use cases or capability gaps, train a focused group (or multiple groups) on those, have them apply the new skills immediately in a real project and then measure and share the outcome. Feed the results and lessons into planning the next cycle and spin again.

For example, imagine each quarter you pick two or three “AI sprints” aligned with business goals. In Q1, the customer support department automates FAQ responses with AI (training agents to fine-tune prompts and interpret the AI’s suggestions), while the finance team tries an AI tool for invoice processing. By Q2, customer support might move on to an AI that helps upsell during live chats and finance might tackle AI forecasting – building on what they learned. Each cycle, more people get comfortable with AI and more use cases get delivered. It’s not overwhelming because it’s broken into bites, but it’s relentless in progressing.

This approach creates a culture of continuous AI learning. Many leading companies now expect employees to dedicate a portion of their time to learning and integrating AI into their work. When you give people permission and time to learn continuously, the skills flywheel really starts humming.

At HCLTech, we encourage such loops through hackathons, “AI days” and ongoing modular training, but the real key is the mindset: always be learning. With each iteration, the workforce not only adds new skills but also new ideas for the business. That developer who learned to use an AI code assistant might next automate a chunk of QA testing, saving time. The marketer who mastered an AI analytics tool might uncover a new customer segment to target. This flywheel effect is how you compound the value of training into innovation.

To sustain it, leaders should regularly celebrate learning milestones just as they do business wins. When your team completes 100 AI skill certifications tied to a project that boosts revenue, acknowledge both the training and the result. This positive reinforcement cements the culture. It tells everyone: this is who we are now – a company that learns, adapts and innovates continuously.

Frontier Firms: The payoff of getting it right

By now, some readers might wonder: is all this skilling and culture work really worth it? The answer from the field is an emphatic yes. The organisations that heavily invest in upskilling their people to use AI – often dubbed “Frontier Firms” for being ahead of the curve – are reaping remarkable rewards. According to a recent IDC study, companies using AI effectively are already seeing an average return of about $3.70 for every $1 invested, with the top performers achieving over $10 for every $1. Those are astonishing figures that underscore a massive competitive advantage. We’re talking efficiency gains, cost reductions and new revenues that far outstrip the initial spend.

Likewise, productivity research out of MIT offers a concrete example: when software developers were given an AI coding assistant, their output went up by 26% on average. Less experienced devs finished tasks much faster and even seasoned ones saved time on routine work. Think about that: a quarter more work done, just by smartly pairing people with AI. And it isn’t limited to coding – the researchers noted similar potential in many knowledge roles. It shows that with the right skills and adoption, AI can significantly amplify human productivity.

Frontier Firms treat these outcomes not as lucky wins but as repeatable and scalable. They realise that getting there requires aligning technology with people from the outset. They back their AI visions with investments in training, change management and incentives for innovation. Culturally, they normalise time for learning – it’s as expected as turning in your weekly report. And crucially, they lead from the top. Executives at these companies don’t just mandate AI adoption; they personally engage with it, talk about their own learning journeys and reward team efforts in that direction. The results speak for themselves in the ROI and productivity data.

And as Vinutha Rao emphasised:

It’s about how we actually look at the entire business point of view, how we sustain and grow our business in the world of technology disruptions and the landscape of skilling. Vinutha Rao

The takeaway is clear: the people and culture side of AI is the real differentiator. The tools are increasingly accessible to all, but how you use them (and who you empower to use them) sets you apart. If you want to join the Frontier Firms at the leading edge, start by looking not at what AI can do, but what your organisation can do with AI — given the right knowledge, mindset and support.

The call to action

The age of endless pilots is over. As executives, we must step up to turn AI pilots into pervasive productivity. This starts with leading by example and fostering a culture where continuous learning is the norm. Make AI skilling a boardroom topic and a standing agenda in team meetings. Put real budget and recognition behind training efforts, just as you would for a major product launch.

As Justin Slade put it:

The future skilling today needs to be always-on, AI-powered, outcome-driven and deeply personalised. Justin Slade

Don’t let your next proof-of-concept become just another tech demo — make it the spark that ignites real business change. Bring everyone into the journey, from the IT developers to the front-line employees who will ultimately use the AI. When people see leaders not only sponsoring but also participating in AI learning, it sends a powerful message that this is serious and here to stay. So, start today. Pick one high-potential project that’s been stuck or slow and apply these principles. Assign an executive sponsor and a clear success metric. Set up a multi-disciplinary team and give them air cover to innovate freely.