The Automation Mirage: Lessons from RPA That Should Terrify the AI Optimists

A few years ago, Automation or Robotic Process Automation (RPA) was hailed as the future of work. It promised to liberate humans from the rigmarole of managing spreadsheets and screens, replacing humans with magic software “bots” that never took breaks or made typos. The vision was irresistible for CEOs: frictionless automation of mindless work, neatly wrapped in a low-code interface and a vendor invoice.

Today, the RPA circus has faded into the background. It is still used and is occasionally useful, but no longer the silver bullet it was sold as. In its place, another acronym has taken centre stage: No points for guessing what it is - AI. Specifically, Agentic AI, the buzzword in boardrooms and brochures alike. CEOs are vibe coding or using LLMs on flights and insisting the whole company follow suit.

But is this new wave of artificial intelligence truly different? Or are we watching a familiar story unfold, one of overpromise, half-execution and strategic positioning by the usual suspects?

The Bad RPA Hangover: Where the Bots Went Wrong

RPA became mainstream around 2016–2018 with the elegance of a well-timed magic trick. It didn’t require ripping out old systems. It sat quietly on your desktop and mimicked a human clicking through screens. It could reconcile invoices, transfer data between applications or process claims faster than a junior analyst.

For businesses still traumatised by expensive ERP rollouts and failed digital transformations, RPA felt like a miracle. The market exploded. UiPath, Blue Prism and Automation Anywhere became billion-dollar names. IT Services and Consulting firms flooded the market with glossy decks and “Centres of Excellence.” Analysts couldn’t publish forecasts fast enough.

And yet, under the surface, there were cracks forming.

While many companies did achieve savings—often between 30–60% for specific processes—few reached any sort of transformative scale. One 2022 industry survey revealed that only 3% of organisations successfully scaled RPA enterprise-wide. Even worse, up to 50% of initial projects were considered failures, often abandoned when bots broke at the touch of a backend system or processes changed or results underwhelmed.

“The problem wasn’t the tech,” one of our clients, an automation lead at a financial services company quipped over coffee…“It was the hype. You had business teams automating processes that shouldn’t exist in the first place. Or worse, automating chaos.”

As it turns out, RPA’s strength and its ability to mimic human inputs was great at some processes but also became its Achilles’ heel when sophistication crept. It could only handle structured, predictable tasks. And when it broke, it broke badly. Bots needed as much maintenance as interns, but with none of the upside: e.g. growth potential, intelligence, perspective and leadership qualities.

What’s more, many outsourcing giants were at least early on, quietly unmotivated to push RPA aggressively. Why? Because every automated task meant fewer billable hours. We know because we’ve advised on so many term papers with automation built in where defining specific processes to be targeted for automation in a contract was always met with resistance. We used the statement - “It was like turkeys selling Christmas.”

Eventually, even the biggest IT firms realised they had to embrace automation or be outflanked. But the shift felt more like containment than conviction.

Enter AI: The Sequel That Could Be Different - Or Not!

Fast forward to 2023. A chatbot writes a university essay. Microsoft integrates OpenAI into Office. Suddenly, every company with a strategy deck and a pulse is talking about artificial intelligence.

The momentum feels unstoppable. PwC estimates AI could contribute $15.7 trillion to the global economy by 2030. McKinsey says generative AI alone could automate activities that account for 60–70% of today’s work hours. Goldman Sachs predicts it could expose 300 million jobs to some form of automation.

From supply chains to legal reviews to customer service, AI is being spruiked as the WD-40 for business friction. No process is safe. No industry untouched.

And unlike RPA, AI isn’t just about mimicking tasks. It’s about rethinking how work gets done altogether. Robotic machines don’t just click buttons; Agentic AI generates content, analyses strategy, even writes and QAs code. If Automation/RPA was a glorified macro, AI is being marketed as a sprightly and enthusiastic office colleague that never says NO.

But here's the catch.

According to a McKinsey 2024 survey, nearly every company is now investing in AI, but only 1% describe themselves as AI-mature - meaning they’ve actually embedded AI across core business functions at scale. Another BCG report notes that 74% of companies struggle to scale AI initiatives beyond early pilots. Sound familiar?

And the obstacles? They're not technological! (Refer to our client statement above)

They’re people. Process. Culture. Leadership.

Many firms simply don’t have clean, structured data to feed their AI models. Others lack the in-house talent to build, train or govern them.

The Service Provider Dilemma: Cannibalise or Cling On?

Sometimes, we feel sorry for the service providers At the centre of this unfolding drama sit the same firms who once pushed RPA: the global consulting and IT services giants. They’re back now, dressed for the AI party.

Many have committed billions to AI services. They’ve partnered with everyone from Nvidia to Anthropic. They’re retraining thousands in generative AI. But the vast majority of the work many of these firms execute still rely on headcount-based billing. And AI, if done right, threatens to obliterate the very thing they sell : people!.

To their credit, some firms are biting the bullet. A few of them can see the writing on the wall and have laid off thousands of staff in less automated lines of work while doubling its AI team. They’ve internal AI copilots to reduce effort in delivery. But the shift is uneven. At many firms, the slide decks are ahead of the delivery muscle.

And then there’s the capability problem.

A lot of big firms aren’t built to execute AI at scale. They’re good at defining strategy, spotting use cases and building prototypes. But when it comes to integrating AI into legacy IT systems (with decades of technology and architectural nuances), training models on messy data - they often rely on specialist talent or the client’s own business and IT teams.

Which means: lesser effectiveness.

So, Who’s Actually Winning?

Despite the fog of hype, a few lessons are emerging.

Firstly, companies that treat AI as an IT upgrade will fail. The winners are treating it as an organisational transformation - investing not just in data and models, but in talent, process redesign and capability transfer.

Secondly, firms that blindly outsource their AI strategy and implementation to vendors will be disappointed. Success requires internal ownership. At ValueKnox we tell our clients “You can’t outsource your future.”

Thirdly, both IT and consulting firms must evolve. The days of selling discrete Data and AI tech, AI and Tech partnerships, operate/BPO pitches are numbered. Clients now want partners who can advise, build, run and teach in the trenches working alongside their teams. Firms that stick to their old models risk becoming irrelevant, no matter how many alliances they sign.

And finally, the opportunity is real. But so is the Hype. The difference between AI and RPA isn’t just the acronym. Done right, AI can create step-changes in productivity. But mind the hype too. Because if organisations don’t do the hard yards : fix their data, empower their teams, kill their silos and pick the right partners – we may be writing about this in similar vein, 10 years from now.

Got thoughts on this article? Let us know in the comments. Share the content if you like it.

Sources:

EY (2019). Get Ready for Robots.

Xenith (2024). Top 4 Reasons RPA Projects Fail

McKinsey & Co. (2022). State of AI 2022: Adoption and Plateau

McKinsey & Co. (2023). State of AI 2023: Generative AI’s Breakout Year

McKinsey & Co. (2024). State of AI in Early 2024

McKinsey & Co. (2025). AI in the Workplace – 1% AI Mature

BCG (2024). Reshaping Business With AI – Survey

Accenture (2024). Reinventing Enterprise Operations with Gen AI

Virtasant (2024). AI Operational Efficiency Case Studies

Infosys (2022). Comment on automation and hiring

HFS Research (2015). Service Providers and RPA

Innovation Leader (2025). The End of Consulting As We Know It

Accenture (2023). Accenture to Invest $3B in AI

Industrial Bank of Korea RPA case – UiPath blog

JPMorgan COIN case – as reported by CEO in media