How HCLTech seeks to bring sanity to AI adoption in a world of tech ‘cosplay’

Ashish Gupta is Head of EMEA for HCLTech and sees first-hand how internal reluctance, driven by uncertainty and minimal knowledge, is holding organisations back from moving away from traditional business models.

He is in a good position to chart the pace of gen AI adoption in the workplace throughout his region. He and his team saw the launch of Chat GPT as a pivotal moment when lots of Proof of Concept (PoC) projects were launched. While AI itself has been around and in use for decades, gen AI has really catalysed the market opportunity to transform business, but are business leaders ready to align the adoption of this technology with their business objectives? I spoke with Gupta to get his perspective.

If your only tool is a hammer, every problem looks like a nail

Machine Learning and automation, along with AI workflow continue to play important roles in enterprise technology, as their part in business transformation is not eradicated by gen AI. Gen AI and agentic AI provide a different role in transformation as they are useful in slicing across business processes. In the agentic area, the ability to create code and apply it to both structured and unstructured data is important, while agentic workflow can change the way that work is done by deterministically creating a co-worker to do a portion of the task.

However, Gupta thinks that leaders have to have an open mindset in order to get the most out of these tools. He explains:

The Board in the enterprise is hearing about gen AI, and the technology fashion show at Davos World Economic Forum plays a role here. This event really cosplays technology as high fashion. One year it is bitcoin that will change the world, the next it is gen AI. Businesses get too taken in by the hype. Our role in the tech ecosystem is to bring sanity to the adoption to enable competitive differentiation. In the UK there are currently a high number of PoCs, but also a lot of confusion.

To address this confusion, HCLTech has its AI Labs to try out use cases with customers, he says:

For example, in the US we discovered that a hospital chain was struggling with surgery time because of a lack of support staff rather than because of a lack of surgeons. It is possible to train agents to perform support tasks in surgery.

The company also has AI Labs in Singapore, Munich, London, and Noida, India where it is seeing increasing enterprise footfall.

Gupta also thinks that the adoption of Microsoft’s CoPilot is adding to the AI confusion, because, as he puts it:

Adoption of MS CoPilot is large but a lot of people are asking about cost and whether adoption is producing productivity gains. Is it generating extra downtime for employees or does it contribute to the P&L balance sheet? There is no answer to that yet.

Deployment need not become a battleground

However, with the early adopters within its client base HCLTech is beginning to gather the data to address these questions. Talking about one of HCLTech’s customers, Gupta explains:

One of our very early adopters is a large law firm in the UK. It moved fast because it felt that the technology could render the practice bankrupt. The fear came from the recognition that in a clearly defined area the technology could do 70-80% of the work. So, the managing partner had the vision to use the technology to amplify the impact of the firm’s lawyers.

Last year HCLTech launched a co-worker para-legal for that law firm which is being adopted at scale. The gen AI agent was trained on all their legal cases, combined with publicly available data. This is an example of a PoC going through to deployment. Adoption is easier in this area because legal frameworks are extremely well-defined and that enables training of AI agents.

The cultural approach for this use case was very simple – the CIO started working with us to create a typical gen AI assistant with a couple of users and this achieved very good outcomes. And so, the gen AI assistant was launched in the rest of the firm.

By starting small and gradually involving more staff, deployment did not become a battleground as the staff had the option whether to use the agents or not. Gupta continues, saying:

However, in many other projects there is a highly restricted use of gen AI because if you democratize agents in the enterprise, they can suck all your data out and so enterprises say to block it. This is one reason why CoPilot adoption is small – there is both the issue of cost and of the need to define rules for employees.

As regards his own experience of gen AI, Gupta adopts it in his sales process, using big data and a predictive agent to identify potential opportunities in order to direct his sales team. He adds:

We are very transparent with customers – if the technology reduces costs by 40% we will share that with the customer by delivering efficiency gains and then we can use that money to do more work in the account.

My take

In HCLTech’s recent end of year results call, CEO C Vijayakumar said that the company expects its business to move, over time, to a mix of 50% people and 50% agentic resolutions, adding that worker location will thus become less important for 50% of the work. Gen AI does pose interesting questions regarding offshoring and the globalisation of labor. It will certainly enable services companies to fast-track their non-linear headcount growth strategies as revenue increases. Whether customers are happy for services companies to determine how agentic savings will be managed in their accounts is less predictable.