‘Agentic AI Companies Are a Cross Between McKinsey and Infos…

Over the last year, the internet has been flooded with confident predictions about AI agents replacing entire departments, automating complex workflows, and reshaping how companies operate. Yet, the impact of these agents is barely visible in most enterprises’ financial statements. 

Some companies, for instance, reporting $100 million in Annual Recurring Revenue (ARR) with AI end up paying $120 million in OpenAI or Anthropic API bills – resulting in losses overall.

The technology exists, but the transformation, more importantly the money it was supposed to trigger, remains elusive. 

Mukesh Bansal, founder of Nurix, CureFit, and Myntra, pointed out the disconnect between the AI agent hype and its actual production deployment in a recent LinkedIn post.

“Everyone is building AI agents and yet so few AI agents are in production doing real work. LLMs keep getting better and better and yet P&Ls are not changing by even basis points! What’s going on?” Bansal wrote.

For him, the reason is simple. Building real, enterprise-grade AI agents is not the same as coding an app or selling a SaaS tool. “Agentic companies are not like a traditional software or platform company at all. It is a cross between McKinsey and Infosys,” he said.

In other words, this would mean bringing together the strategic consulting and problem-solving expertise of a firm like McKinsey with the large-scale technology implementation and delivery capabilities of a company like Infosys. Through this, agentic AI companies can essentially combine high-level business strategy with the ability to execute it through technology at scale, instead of just building agents and providing them to customers.

Lesson for Agentic AI Companies

Bansal said that companies have to fundamentally transform business processes and build agents, not just automate old processes as that would be bad for business.

The challenge is as much human as it is technical. “Building enterprise grade agents requires the best of human talent and LLMs working together. One has to reimagine the future and then build agents that can dynamically evolve,” he said.

His analogy of McKinsey plus Infosys struck a chord across the industry. Raoul Nanavati of Navana.ai called it “an elegant way to put it” while Revathi S, founder of Version Best, simply said, “finally, someone said this!”

For Varun Krishna, founder of The AIGency, the gap is clear in marketing. “One of the hurdles [in marketing] is overcoming human resistance to fully leverage, train and push these agents to handle core parts of the marketing process which will eventually get resolved.”

In an earlier conversation with AIM, Bala Prasad Peddigari, chief innovation officer for technology, software services business group at TCS, said that the current AI agents are not yet accurate enough to be built for production. But he called it an opportunity.

In his assessment, more than 90% of Indian companies are currently dabbling with AI in fragmented ways, lacking both strategic cohesion and scalable frameworks.

Read: TCS Innovation Chief Thinks Accuracy in Agentic AI is Still a Question Mark

Full Stack AI Companies

Neeraj Sinha, who is building an early-stage startup, believes the answer lies in what he calls “full stack AI companies” rather than toolmakers. 

“We should expect full stack AI companies to show sizable EBITDA benefits. Some problems will continue to be better solved using deterministic programming techniques while others can be better solved with traditional ML and yet others with genAI or agentic,” he said.

Most companies do not seem ready for it.

According to a recent report by Capgemini, only 2% of organisations worldwide have fully scaled AI agents. While 78% of executives say they have launched GenAI initiatives, nearly the same percentage admit they have yet to see any meaningful bottom-line impact — a clear sign of the emerging “GenAI paradox.”

Dr Vikram Singh, head of AI at Mahindra, said most AI deployments so far have targeted low-hanging fruit. “Most current AI implementations have focused on automating routine, day-to-day tasks and replacing entry-level support roles. While this has yielded some cost savings, these reductions are often modest compared to the initial investment.”

The infrastructure gap is equally stark. According to a report by StackAI, 85% of technology leaders say they need infrastructure upgrades before they can deploy AI at scale. 42% of enterprises require access to eight or more data sources for AI agents to be effective, yet fewer than 20% say their data is ready for such integration.

The gap between a quick prototype and a production-ready system is wider than many care to admit. 

The Flip is Also True

Ashay Tamhane, founder of WhatsYum, believes many projects start backwards. “Most are starting with a solution instead of starting with a business problem. Investors are running behind solutions because of FOMO. Customers still want real humans to talk to. It’s a funny world.”

To cite an example, most consulting firms are earning a lot of revenue from generative AI alone. In FY2024, Accenture raked in $900 million from generative AI, up 9x from the previous year, and ended with $3 billion in bookings. This year so far, it just reported $1.5 billion in generative AI bookings.

ServiceNow, meanwhile, says its AI product Now Assist will hit $1 billion in ACV by 2026. 

OpenAI seems to have learnt this lesson. It is no longer content with just powering those systems in the background. According to recent reports, OpenAI is courting large enterprises and government clients, ramping up its consulting muscle. The AI company achieves this by embedding its own engineers and researchers to tailor AI systems for specific business needs, adopting a strategy similar to that of Palantir, Accenture, and ServiceNow.

Read: Accenture, ServiceNow and Palantir Teach Startups How to Make AI Money

That is indeed the right path forward for agentic AI companies to be consulting companies.

Prakash Sharma, senior manager, digital transformation strategy at Lenovo, pointed to a Microsoft statistic showing just 4-5 agents per organisation on average, most still in the proof-of-concept stage. “The key reason is that orgs are not ready with AI data and knowledge so currently only around 5-10% of AI POC move to production,” he said.

In the end, the consensus among practitioners is clear — building a true agentic company requires more than prompts and APIs. It demands rethinking processes from the ground up, blending strategic vision with deep technical execution, and having the patience to push through the resistance that comes with any major organisational change.

Or as Mukesh Bansal put it, if this was easy, agents would already be running the world.

The post ‘Agentic AI Companies Are a Cross Between McKinsey and Infosys’ appeared first on Analytics India Magazine.

Generated by RSStT. The copyright belongs to the original author.

Source