A comprehensive Q&A from GrowthSutra's LinkedIn Live session featuring host Vishwendra Verma (Founder & CEO, GrowthSutra) with industry experts Dharmendra Kapoor (Former CEO at Mindsprint and Venosa), Praveen Savant (Independent Advisor, Global Board Governance and Technology Advisory).
“In 2025, AI startups captured 51% of all global venture capital – about 310 billion dollars – most of it into companies that didn’t exist five years ago. Meanwhile, Indian IT services giants – TCS, Infosys, Wipro, HCL, Cognizant – collectively lost 150 billion dollars in market value in just nine months. TCS, which had never laid off people in 57 years, cut 12,000 jobs in July, and Accenture is down 25% over the last year.”
“At the same time, 77% of enterprises now expect their IT partners to deliver AI expertise, not just labor arbitrage, while 95% of AI pilots never make it to production. So the ground is clearly shifting.”
“What happens when the ground beneath an entire industry disappears? Are we moving too fast, too slow, or asking the wrong question entirely?”
What’s Really Changing on the Ground?
Q (Host – to Dharmendra):
“DK, you wrote, ‘AI is not a bubble, but unchecked AI hype might be.’ Are we investing ahead of proven returns? Are we moving too fast, too slow, or just asking the wrong question?”
Dharmendra Kapoor: “Technology has gone through similar cycles: ERP, dot-com, remote infrastructure, cloud, digital – roughly every eight years there is a disruption, a narrative that ‘this will change everything,’ and yet every time we eventually find new opportunities and see businesses expand through tech. AI is no different in that sense.”
“The twist this time is that even IT service providers themselves are being disrupted, because a lot of work they used to do is now getting automated by AI. That makes the disruption feel much closer to home, but it also means we must reshape how services are delivered, not just rethink them.”
Q (Host – to Praveen):
“Praveen, from the buyer’s side, how do you see this landscape? What’s really changing in buying behavior?”
Praveen Savant: “From the buyer’s lens, the buyer holds the risk, so outcomes matter more than ever. Earlier we outsourced coding; now with AI we’re effectively outsourcing cognition – and that demands governance and ethical models.”
“I’ve seen dozens of experiments over the past 8–9 years. Many Indian AI initiatives that had great technology – like an India-specific LLM or secure communication platforms – failed not for technical reasons but due to distribution, culture and adoption challenges. At the same time, there are Indian healthcare AI platforms now running successfully in 70+ countries. So there are both setbacks and solid wins, but what’s clearly changing is that governance, accountability and ethics are no longer optional.”
He adds that alongside GDPR, India’s DPDP law is an added layer every AI-heavy or agentic solution must comply with, which increases effort and scrutiny.
How Client Conversations Are Evolving
Q (Host – to Dharmendra):
“From your experience leading and transforming services organizations, how have client conversations evolved compared to 2–3 years ago? Where are firms struggling to pivot?”
Dharmendra Kapoor: “Client conversations today are dominated by: ‘How do we experiment with AI?’ and ‘Where can we show quick returns?’”
He describes two client archetypes:
Early adopters: Want competitive edge and are willing to invest in AI experiments, even for small leads.
Followers: Spend less on experimentation, wait for proven use cases and ask for ‘what’s working elsewhere that I can adopt cheaply?’
COVID accelerated this urgency; companies that adopted e-commerce early reopened faster and grew faster, creating a FOMO effect that is now carrying into AI adoption.
“Clients are asking both linear questions – ‘How can I improve efficiency, reduce cost, improve productivity?’ – and transformational questions – ‘Can I change my business model or create new revenue streams?’ That’s the exciting part; everyone’s appetite differs, but the direction is clear towards quicker AI adoption.”
Audience Q&A – All Questions and Names
1. Linear Wins from AI in Operations
Audience Comment – Jinchandra Mule:
“We have seen more productivity improvements using automation and AI data analytics – reduced SLA tickets through Duallist implementation, and great knowledge share from KT docs.”
Response – Dharmendra Kapoor:
“That’s a classic example of linear improvements – automation, analytics, reduced ticket volumes, better knowledge reuse. There are many such successful use cases already in production. The next wave will be more complex pilots and projects that aim to change business models, not just improve existing ones.”
2. Revenue Models in the AI Cycle
Audience Question – Bharti:
“How do you see the revenue models changing with the AI cycle (for IT services)?”
Answer – Dharmendra Kapoor:
“For IT services, traditional revenue models won’t vanish overnight. There will still be legacy services for a while. But over time, revenue will shift toward AI-enabled platforms and assets.”
He explains with ERP as an example:
Historically, ERP processes are deterministic – sequential, predictable.
With AI, processes become probabilistic – you give a task or prompt and you get an answer without fully seeing the path.
“That creates room for AI ‘agents’. For instance, an Order-to-Cash (O2C) agent that sits across any ERP, built and owned by a services firm.
The company can then charge for the agent itself – via license or subscription – instead of only billing people-hours for O2C support.”
“If services companies continue to respond with ‘I’ll give you five people to build your agent’, they’ll stay stuck in cost-plus models. To change the revenue model, they must change the business model – become services plus asset organizations.”
Additional Note – Praveen Savant:
He uses the analogy of moving from petrol/diesel cars to electric vehicles – fewer components, different architecture – but warns that we’re not yet at fully autonomous cars without ethical constraints. Reality is somewhere in-between, and services must evolve accordingly.
3. Counterpoint on AI Speed
Audience Comment – Aman Pandey:
“I think you’re wrong. Have you tried building using the latest IDEs? 100 times more work at 100 times the speed and high-quality output.”
Host’s Framing (Vishwendra):
Vish acknowledges that modern tooling dramatically increases speed and quality, and that industry will continue to test new models and revenue architectures as these tools mature.
4. Distribution Muscle for AI Solutions
Audience Question – Prashant Dixit:
“Everyone is rushing to adopt AI in transforming their businesses, functions, teams. But for a solution/services company, distribution muscle is the key. How to build a great distribution network? Would love your thoughts.”
Answer – Dharmendra Kapoor:
He first addresses the supply chain analogy: historically, logistics and distribution have been highly customized with low standardization, but data-driven models have improved visibility and planning significantly.
“With AI, agents can recalculate routes and plans on the fly – considering weather, traffic and constraints – to optimize distribution. These use cases are being actively experimented with in supply chain.”
He then extrapolates to AI solution companies:
Initially, solutions may be built for specific clients.
Over time, as agents are trained on more data, they generalize to broader user sets.
Ultimately, they should be distributed via platforms and marketplaces, not only via one-on-one selling.
Answer – Praveen Savant:
Praveen recalls classic CPFR (Collaborative Planning, Forecasting and Replenishment) implementations at Godrej Consumer Products, and notes that Unilever and Amazon are already using agentic AI to forecast demand and optimize supply chains with thousands of SKUs.
Coming back to AI solution firms, he highlights a critical challenge:
“Some AI products were good, the intent was good – but they lacked distribution muscle. Without presence, connectors into ERPs/CRMs, or co-selling models, it’s very hard to scale. This needs deeper thinking and partnership-led distribution strategies.”
5. Fixing Process Leaks Before AI
Audience Question – Arun Mathew:
“Very few organizations are asking a key question: What are the business process and efficiency leaks? How do you fix the process? Just doing AI or AI agents for the hype is where these implementations fail.”
Answer – Praveen Savant:
“Arun, you’re absolutely right. The first principle must be: What business problem are we solving? Technology comes second.”
He cites an example: under pressure from compliance (DPDP, GDPR, etc.), his team faced huge overload in contract review.
They applied NLP-based AI to automate parts of legal review and brought turnaround down from 8 days to 8 hours, but kept human-in-the-loop to supervise and guard against bias, false positives/negatives, and hallucinations.
“Buying an AI product and then hunting for a problem to attach it to is exactly how you end up in the 95% of failed pilots.”
6. Balancing Services Culture with Product Mindset
Audience Question – Aashil Shah:
“Many services firms are now looking to build their own AI assets or accelerators to compete with product companies. What is the key to balancing a service-oriented culture with a product mindset required to build scalable AI assets?”
Answer – Dharmendra Kapoor:
“It’s not an easy transition. Services and product cultures are fundamentally different, and historically services companies have struggled with pure products.”
“However, AI opens a unique opportunity to build reusable accelerators and agents – for example, an O2C agent – that can be deployed across multiple clients and platforms.”
He suggests two models:
Product-like: Build AI agents and sell them via licenses as products.
Subscription/Marketplace: Host agents on a marketplace, let clients download, configure (with help from the services team), and pay via subscription – a services-plus-assets model.
He also notes that product companies like SAP and major CRMs will build their own agents, so services and product firms will likely compete and co-exist in this new “agents + services” space.
Answer – Praveen Savant (Build vs Buy angle):
He adds that Global Capability Centers (GCCs) are increasingly where AI adoption and innovation are happening, not just at global HQs. GCCs consolidate process knowledge and efficiency initiatives, building in-house AI first, then outsourcing older technologies later once they mature.
7. Pricing vs “PhD-Level” AI
Audience Question – Deepak Mehta:
“When IT companies charge $35/hour, how will they compete with ‘PhD-level’ AI at $20 per month (e.g., OpenAI models)? With experts at ~50% cost and 176 hours a month, this seems like a huge revenue compression. How will IT services survive?”
Answer – Dharmendra Kapoor:
“These are real concerns and are driving a lot of the talent disruption conversations. Earlier, ‘I can code’ was a valuable statement because programming was not automated. Now, much of that is getting automated.”
He uses a construction analogy:
Old role: Someone who lays bricks (writes code).
New role: Someone who designs and orchestrates the building (architect, designer).
“In AI-era IT, human roles will shift from performers to orchestrators and architects. People will set goals, design flows, integrate agents, check results for bias and correctness, and stitch systems together.
Certifications in narrow skills will age quickly; continuous upskilling in AI usage and orchestration is critical.”
Answer – Praveen Savant (Buyer’s view):
“As a buyer, I’m not overly worried by the $20-per-month AI vs $35-per-hour engineer comparison. I don’t buy only on cost. I also buy accountability.”
“ChatGPT or any AI might give an outcome, but who owns that outcome if it goes wrong? With a services partner, I have someone accountable. Until AI solutions also come with clear accountability, governance, and remediation, I won’t pick the cheapest option purely on cost.”
8. Comment on “Service as Software” Pivot
Audience Comment:
“Great thoughts, DK. IBM Consulting CEO Mohammed Ali has also been talking about ‘service as a software’ as a pivot for consulting for some time now.”
The host treats this as a reinforcing comment, aligning it with DK’s view that services will increasingly be delivered as software + agents + platforms, not only as manpower.
9. AI as Human Brain Replica & Cost/Hours
Audience Comment & Question – Biplab Mohanty:
“AI is actually a replica of the human brain and how we train it. Basics are important to implement and design. Will AI reduce the cost and man-hours?”
Host’s Response (Vishwendra):
He acknowledges that productivity gains and cost reduction are real and already observable. But, as Praveen emphasized, ‘human-in-the-loop’ design is still crucial due to risks of bias and hallucination.
Current systems aren’t yet ready to fully address these risks autonomously, so AI will reduce some man-hours but simultaneously create new oversight and governance work.
10. Human Dimension – Talent & Careers
Q (Host – to both, referencing industry stats):
“Reports say 400,000 to 500,000 IT services professionals could face displacement by 2027, with about 70% in the 4–12 years experience bracket. For leaders across services, enterprises and startups, where should they invest in upskilling and talent so they can sustain through this shift?”
Answer – Dharmendra Kapoor:
He highlights four major skill pillars:
Data skills: Managing, cleaning and transforming data so AI can give trustworthy results.
Business process skills: Understanding as-is processes deeply and redesigning them into AI-era probabilistic flows, especially as every application starts to embed agents.
Domain skills: Combining industry/domain expertise with AI so teams can articulate the right business outcomes and craft strong models.
Orchestration skills: Setting goals for agents, validating outputs, understanding the “black box” paths to decisions and managing bias/trust issues.
“These skills must be built inside organizations first, then eventually pushed back into academia via reverse engineering of curriculums. Talent development cannot wait for colleges to catch up.”
11. New Roles Around Context, Risk & Accountability
Audience Question – Prashant Godrehal:
“We believe there are things AI misses: context, risk ability, accountability, purpose, truth checks. How do you value these versus ‘yester-year’ skills?”
Answer – Praveen Savant:
“AI or any technology does not, by itself, reshape careers. It displaces those who don’t evolve.”
He breaks it down by career stage:
Young engineers: Need systems thinking rather than only coding or platform skills.
Senior leaders: Don’t need to learn Python, but must understand governance frameworks, right questions to ask, and what to expect from AI systems.
Mid-level professionals: This is where the biggest opportunity lies – becoming human-in-the-loop supervisors, bias controllers, hallucination checkers, data custodians, DPDP/data fiduciaries, etc.
He frames these as a new flywheel of AI-era roles that directly address what AI currently misses: context, risk, accountability and truth checks.
Closing Thoughts & 2026 Hint
Host (Vishwendra Verma):
“We opened with a provocative question: Is IT services doomed? That was never the right question. The real question is: What does IT services become when AI becomes the operating system of business?”
He notes two contrasting data points:
The market has already punished status quo, with $150B of value erased from legacy models.
At the same time, ‘service as software’ could represent a $1.5 trillion market by 2030, where headcount pyramids invert and senior leaders become AI orchestrators rather than people managers.
Dharmendra Kapoor (on 2026 trend):
“If I had to pick two words for 2026: ‘Great Rewrite.’ We’ll have to rewrite business processes and applications to embed agents. The opportunity will only grow from here.”
Praveen Savant (on 2026 trend):
“I’m very bullish on domain-specific models – they’re already showing great success stories, and that will accelerate.”
Session Date and Recording
This Q&A is based on GrowthSutra’s “202X Vision: The AI Product Gold Rush - Is Your IT Services Business Doomed? - CEO’s Perspective” session held on 17 December. You can watch the full conversation and framework deep dives on GrowthSutra’s YouTube channel to see the complete discussion in context.