McKinsey, BCG, Bain, Deloitte, PwC and more: the 2026 research, summarised and ranked

McKinsey, BCG, Bain, Deloitte, PwC, Accenture, EY and KPMG publish hundreds of reports a year between them. Each one is long, each firm has a house view, and most summaries you find online come from someone selling an AI tool. So we read the flagship 2025 and 2026 publications ourselves and put them on one page: the ten reports worth your time, ranked, what they agree on, where they quietly contradict each other, and what it all means if you run or are starting a business.

A note on neutrality before you read further. We have no relationship with any firm on this page. Nobody paid to be here, nobody was left out for being a competitor, and we quote each firm's own numbers rather than our opinion of the firm. Where a figure only appears in press coverage, we say so.

How we ranked the reports

A report ranks higher if it scores well on four tests:

• Method. How many respondents, across how many countries, and is the method published.
• Freshness. Data collected in 2026 beats data collected in 2025.
• Decision value. Does reading it change what you would build, buy or stop doing? A report that only says AI matters scores low here.
• Independence of the claim. Survey findings rank above modelled market sizes, because a forecast is a scenario, not a measurement.

The one-table view

The 10 reports, ranked, and what each one actually says

The most cited AI survey in business, now close to a decade old, and this edition is the most candid yet. Eight in ten respondents say AI has improved their own productivity and half say it helps them decide better. Yet only 37 percent of organisations report any positive EBIT impact, essentially unchanged from 2025. The "high performers" who attribute 5 percent or more of EBIT to AI are still about 6 percent of the sample.

The agent story is a size story. Forty percent of respondents from companies above USD 1 billion in revenue say they are scaling AI agents, up from 27 percent. Smaller companies are stuck at 22 percent. About one in five say AI running costs, including token bills, now limit how much they use it.

Why it ranks first: it is the baseline every other report argues with. Personal productivity is real; enterprise profit is not yet. That gap is where most of the business opportunity of the next five years sits.

If you build products, this is the single most useful document published this year. Bain looks at the work people do between software systems: copying data from the ERP to the CRM, chasing an invoice, reconciling a payment, following up on a ticket. It sizes the market for AI agents that do this work at about USD 100 billion in the US and roughly USD 200 billion once Canada, Europe, Australia and New Zealand are added. Vendors capture only about USD 4 to 6 billion of it today, so more than 90 percent is open.

Two details matter more than the headline. First, the share of work that can be automated varies a lot by function: roughly 40 to 60 percent of tasks in customer support and engineering, but only 20 to 30 percent in legal, where mistakes are expensive. Second, the work that goes first is work with a result you can verify. Bain also notes pricing is moving away from per-seat licences toward usage and outcome-based models.

Why it ranks second: it does not just say AI matters. It tells you which work, in which functions, and why now.

The biggest sample of CEOs anywhere, and the bluntest numbers on AI returns. About 30 percent of CEOs saw more revenue from AI in the past year and 26 percent saw lower costs, but 56 percent saw neither. Only one in eight got both, and PwC calls that group the vanguard: they built proper foundations and applied AI far more widely, including to their own products.

The rest of the survey is worth reading too. Only 30 percent of CEOs are very confident of revenue growth over the next year, down from 38 percent and from 56 percent in 2022. Cyber risk exposure is up to 31 percent from 24 percent. And 42 percent say their company has started competing in new sectors in the last five years. India also moved up as an investment destination: 13 percent of CEOs planning foreign investment now pick it, up from 7 percent.

Why it ranks third: scale, and the vanguard analysis, which tells you what the 12 percent did differently.

The freshest data on this list, and it asks employees as well as bosses. Seventy-five percent of leaders plan to raise AI spending over the next year. Fifty-three percent expect measurable return within 12 months, but only 25 percent report widespread, sustained value today. The most telling pair: 41 percent of leaders use AI every day, yet only 15 percent have redesigned processes end to end and 20 percent are deploying agents across functions.

The mood is also shifting. Leaders expecting domestic revenue growth fell from 90 percent in January to 71 percent. The share who see signs of an AI bubble dropped from 51 percent to 32 percent. And 69 percent now watch AI operating costs more closely than a year ago. On the employee side, 82 percent say AI made them more productive, while half say it has raised the pace expected of them.

Why it ranks fourth: most recent fieldwork, two-sided sample, and the clearest number for the "use without redesign" problem.

The spending report. Companies expect to roughly double AI spend in 2026, to about 1.7 percent of revenue. Ninety-four percent say they will keep investing even if returns take more than a year. AI has also become the boss's problem: 72 percent of CEOs now call themselves the main AI decision maker, about twice last year's share, and half believe their own job depends on getting it right. Around four in five are more optimistic about AI's return than a year ago, and about 90 percent expect agents to deliver measurable returns this year.

Why it ranks fifth: it explains why demand for AI work will keep growing even while returns lag. The budget is committed and the CEO owns it.

Five trends: AI goes physical (robots, autonomous systems in warehouses and factories), the agentic reality check, the AI infrastructure reckoning, the great rebuild of IT, and the AI dilemma in security. The line most worth remembering is in the agents chapter: many organisations are piloting agents, few have them in production, and failures often come from automating a broken process instead of redesigning it. Deloitte also notes that falling cost per token is being outpaced by rising usage, which is pushing companies toward hybrid cloud, on-premise and edge setups.

Why it ranks sixth: a clear map of the technology layer, but no survey data, so it informs more than it proves.

The biggest number on this page. McKinsey estimates AI agents that shop on a customer's behalf could orchestrate around USD 1 trillion of US retail revenue by 2030 and USD 3 to 5 trillion globally. The estimate covers goods only and is built on US Census data and adoption forecasts. Today the behaviour is small: about 15 percent of consumers have used AI tools on retailer sites and only 1 to 3 percent have bought through an AI platform. McKinsey's advice to merchants is practical: make product data, APIs and fulfilment signals readable by agents before building any agent of your own.

Why it ranks seventh: the direction is right and important, but a modelled number. Other forecasters publish far smaller figures, so treat the range as a scenario, not a fact.

Not a consulting firm, but the reference every consulting firm cites on work. By 2030, 170 million jobs are created and 92 million displaced, a net gain of 78 million. Job disruption equals 22 percent of today's jobs, and nearly 40 percent of the skills needed on the job will change. Out of every 100 workers, 59 will need reskilling and 11 of them are unlikely to get it. Frontline, care and education roles add the most jobs; AI, data and renewable energy specialists grow fastest. Seventy-seven percent of employers plan to upskill, and 41 percent plan to reduce headcount where AI automates tasks.

Why it ranks eighth: essential context for hiring and training decisions, though its data is now nearly two years old.

EY's CEO Confidence Index fell from 83.0 to 78.5. Fifty-eight percent of CEOs expect AI to be a major growth engine within two years, but only 20 percent say it significantly beat expectations last year. The jobs signal is softer than many headlines suggest: the share of CEOs expecting AI-driven headcount cuts fell from 46 percent in January 2025 to 24 percent in December 2025. Partnerships are back too, with 79 percent planning joint ventures or alliances in 2026, up from 62 percent.

Why it ranks ninth: solid quarterly pulse, useful as a check on the bigger surveys.

KPMG's framing is that generative AI has moved from future ambition to present reality, shaping strategy, margins and workforce design at the same time. Leaders are retooling teams for people and AI working together, and 39 percent of US CEOs name better end-to-end supply chain visibility through data and AI as their top resilience move. A smaller KPMG pulse of 100 US CEOs, reported by Axios, found only 9 percent plan to cut staff because of AI this year, while 55 percent expect to hire more.

Why it ranks tenth: a credible sample, but the public summary carries fewer hard numbers than the others.

The India shelf: four reports to add if you work in India

• BCG and FICCI, India's Triple AI Imperative (December 2025). AI adoption in India's MSMEs could unlock over USD 500 billion of economic value by 2030, through productivity, lower costs and better access to credit. The barriers it names are digital infrastructure, awareness and skills. Report PDF
• Bain, India Enterprise Technology Report 2026. AI and data work is expected to take 40 to 45 percent of Indian enterprises' change-related tech spend in 2026, with a clear gap between investment and value delivered. Read it
• Bain and IVCA, India Venture Capital Report 2026. Indian VC reached roughly USD 16 billion in 2025 and investors are now in a profit-and-exits mood. Conviction themes include AI, climate and deeptech. (Figures from press coverage.)
• BCG consumer research on GenAI in India. Around six in ten Indian consumers already use GenAI tools, one of the highest rates in the world, and it is changing how they research purchases. (From press coverage.)

Six things every firm agrees on

1. The AI value gap is the market. Usage is close to universal. Profit is not. McKinsey's 37 percent, PwC's 12 percent vanguard and Accenture's 25 percent with sustained value are three ways of saying the same thing.
2. Redesign beats adding tools. BCG, Deloitte, Accenture and PwC all land here. The winners rebuild the workflow around AI. Bolting a chatbot onto an old process rarely moves profit.
3. Agents are the next spending wave. BCG says over 30 percent of 2026 AI investment targets agents. McKinsey sees large firms scaling them fast. Deloitte warns most are still pilots.
4. Checkable work goes first. Bain's rule: if the output can be verified (a matched invoice, a closed ticket, a payment received), it gets automated first. Judgement-heavy work with expensive mistakes comes later.
5. Pricing is shifting. From per seat to per use and per outcome. That lets a new entrant compete on how much risk it takes off the customer rather than on price.
6. Trust, security and skills are the new constraints. PwC's cyber exposure jump, Deloitte's AI dilemma chapter, the WEF skills gap and McKinsey's cost constraint all point the same way.

Where they quietly disagree

Reading the reports side by side shows three tensions that no single report mentions.

• Optimism versus evidence. BCG finds four in five CEOs more optimistic about AI's return and about 90 percent expecting agents to pay off this year. PwC, from a larger sample, finds most have not seen any return yet. Both can be true: confidence is running ahead of results. That gap is a risk for anyone selling AI on promise alone.
• Jobs. The WEF found 41 percent of employers planning to cut staff where AI automates tasks. A year later, EY found only 24 percent of CEOs expecting AI-driven headcount cuts and a KPMG pulse found 9 percent. The likely reading: companies expected cuts, then found AI changes tasks more than it removes roles, at least so far.
• Market size. McKinsey's USD 3 to 5 trillion for agentic commerce and Bain's USD 100 billion for cross-system agents measure different things in different ways. Do not add them up, and do not put either in a pitch deck without the definition next to it.

What this means if you run a business in India

• Before buying any AI tool, pick one process and redesign it. One measurable workflow, end to end, beats ten tools used casually. Our guide on what to automate first ranks the usual candidates by payback.
• Measure in money, not in usage. The firms all say the same thing: teams feel more productive long before the P&L moves. Decide the rupee number before you start.
• Get ready to be chosen by AI, not just found by Google. Six in ten Indian consumers already use GenAI, and AI shopping agents read structured data, reviews and citations. This is the work we cover under AI search optimisation.
• Treat security as a growth issue. Cyber risk now sits next to economic volatility as the top CEO worry in PwC's survey.
• Upskill before you hire. Seventy-seven percent of employers plan to upskill. For a small business, one trained person who owns your AI workflows is worth more than a new tool subscription.

Ten business opportunities the research points to

We asked one question of all this research: where do a growing need, an underserved customer and a checkable result overlap? Below are ten ideas, scored out of 30 on demand, revenue potential, scalability, defensibility, speed to market and how buildable they are for a small team. They lean toward India because that is the market we know, but most travel to any economy full of small businesses on WhatsApp.

The three strongest

1. Vernacular voice agent for lead callbacks (25/30). Calls every new enquiry within a minute in the caller's language, qualifies it, books the visit and updates the CRM. It fits Bain's verifiable-output rule: the appointment was booked or it was not. It rides India's voice-first behaviour, with around six in ten consumers already using GenAI, and most local businesses still lose paid leads to slow callbacks. Fastest to revenue and easiest to sell, but crowded, so the moat has to come from vertical scripts, language quality and the data each conversation adds. Model: base fee plus usage per connected minute.

2. Agentic commerce readiness (24/30). Audits how AI assistants see a business, fixes product feeds and structured data, monitors citations, and connects to agent checkout protocols as they arrive. It rides McKinsey's biggest trend. The risk is well-funded global competitors, so the opening is the Indian SMB tier they ignore.

3. Vertical operating system for gold finance (23/30). Gold buyers, pawnbrokers and small gold-loan lenders still run on paper, WhatsApp and old desktop software, while KYC, anti-money-laundering records and lead handling stay manual. Tighter gold-loan rules keep pushing this market toward proper software, and competition is mostly small legacy vendors. Model: annual licence per branch. Disclosure: we build software for this market ourselves, so weigh this one knowing that.

The single strongest

On balance, the vernacular voice agent. It has the broadest demand (every business that pays for leads), the fastest route to a first paying customer, a result the customer can count every week, and a market where English-first tools leave most callers underserved. The honest caveats: it is a crowded field, so a team needs a clear vertical and language focus from day one; India's telecom rules (DLT registration, and the newer rules on AI calls) need setting up early; and the agent should always say it is an AI. A two-week trial against a human control group should come before any serious build.

How to read a consulting report without being sold to

Consulting research is some of the best free material in business, and it is also marketing for the firm's services. Five habits keep you on the right side of that line.

• Check who was asked. A survey of companies above USD 500 million in revenue tells you little about a 20-person business. Read the methodology box first.
• Separate surveys from forecasts. "37 percent of respondents said" is a measurement. "USD 5 trillion by 2030" is a model with assumptions.
• Read the same question across firms. When McKinsey, PwC and Accenture all find the same profit gap with different methods, believe it. When only one firm finds something, wait.
• Find the "so what" for your size. Most reports are written for large company boards. Translate each finding into one decision you could make this quarter.
• Notice what is missing. Few reports cover India's small businesses directly. That blind spot is itself an opportunity.

About this roundup

Apex Influence is a growth firm in Bangalore, and we are building our strategy practice in the open. Reading the major research carefully is part of how we advise clients, so we publish the reading. We will update this page as new editions come out. If you spot a report we should add, or a number that has changed, tell us and we will check it. You can also see how we approach planning on our marketing strategy page.

Frequently asked questions

Sources

1. McKinsey, The state of AI in 2026: On the road to ROI (Aug 2026): mckinsey.com
2. Bain, The USD 100 billion SaaS opportunity hiding in cross-system labor (May 2026): bain.com
3. PwC, 29th Global CEO Survey (Jan 2026): pwc.com
4. Accenture, Pulse of Change (Oct 2026): accenture.com
5. BCG, AI Radar 2026 press release (Jan 2026): bcg.com
6. Deloitte, Tech Trends 2026 (Dec 2025): deloitte.com
7. McKinsey, The agentic commerce opportunity (Oct 2025): mckinsey.com
8. World Economic Forum, Future of Jobs Report 2025 (Jan 2025): weforum.org
9. EY-Parthenon, CEO Outlook Survey press release (Jan 2026): ey.com
10. KPMG, 2026 CEO Outlook: kpmg.com; KPMG CEO pulse coverage: axios.com
11. BCG and FICCI, India's Triple AI Imperative (Dec 2025): bcg.com PDF
12. Bain, India Enterprise Technology Report 2026: bain.com
13. Bain and IVCA India VC Report 2026 coverage: businesstoday.in
14. BCG GenAI and Indian consumers coverage: tribuneindia.com
15. MSME formal credit access coverage: newkerala.com

Keep reading