Salesforce Puts Agentic AI ROI Math in Focus

Salesforce ties India's growth case to autonomous agents

Salesforce said agentic AI could add US$500-600 billion to India's GDP by 2035, People Matters reported on September 16, 2026. The company highlighted the research at a media briefing on the sidelines of Dreamforce 2026, which People Matters attended. People Matters reported the Salesforce projection; the figure has not been confirmed elsewhere.

India is moving rapidly from AI experimentation to implementation, Salesforce said at the briefing, with organisations increasingly deploying autonomous AI agents across customer service, operations and business workflows. Gavin Barfield, Vice-President and CTO, Solutions, ASEAN at Salesforce, was one of the executives People Matters named as speaking at the briefing.

Salesforce positioned Dreamforce 2026 around the concept of the agentic enterprise, describing a future in which humans and AI agents work together using trusted data, applications and workflows, according to People Matters. Enterprises are looking beyond generative AI experimentation and focusing on operational use cases capable of delivering measurable business value, company executives said during the briefing.

The India research reports deployment, payback and satisfaction

Salesforce's latest India-focused research covers deployment rates, payback time and customer outcomes, People Matters said. Adoption is delivering results faster than many organisations anticipated, the company said at the briefing. Improvements in customer experience outcomes among adopters also appear in the study, according to People Matters.

Findings for business leaders in India sit alongside a separate set of percentages for Global Capability Centres, People Matters reported.

Nilekani told a Mumbai fintech audience small firms will hire

Infosys cofounder Nandan Nilekani said startups and small businesses, rather than large corporations, will be the main source of jobs as AI spreads through the economy, Entrepreneur India reported on September 14, 2026. He spoke at the Global Fintech Fest 2026, held at the Jio World Centre and Trident, Bandra Kurla Complex, Mumbai.

Large companies with structured roles and standardised processes are better placed to deploy AI for automation, which could shrink their workforce, Nilekani said at the event. Smaller businesses with more varied roles are harder to automate and will be the bigger job creators, he added, per Entrepreneur India. Both new-age startups and traditional small businesses will be the employers of the future, he said.

The seventh edition of the festival carried the theme Potential to Impact. Its three focus areas were agentic AI, tokenisation and quantum computing, and it drew policymakers, regulators, bankers and fintech founders from over 80 countries, Entrepreneur India said.

Tokenisation paired with AI agents could give small businesses access to capital and markets currently available mostly to large enterprises, Nilekani said. He said the Reserve Bank of India is exploring tokenised certificates of deposit and working with SEBI on tokenised corporate bonds. Three pilots built on the Finternet architecture cover cattle records, warehouse receipts and loan documents standardised with smart contracts, Entrepreneur India reported.

McKinsey published a trends outlook; People Matters reviewed monitoring

McKinsey published its Technology Trends Outlook 2026 on September 16, 2026, describing it as an annual analysis that ranks the top tech trends that matter most for companies. The firm frames the outlook around which new technology will have the most impact in 2026 and beyond.

People Matters published a review of workplace monitoring on September 16, 2026, reporting that surveillance moved from an emergency management tactic during the 2020 and 2021 expansion of remote work into a mainstream practice. Vendors promised productivity insights, accountability and better workforce management, that review said. Monitoring tools now track screen activity, keystrokes, location data and communication patterns, People Matters reported.

A 2021 survey reported by HRMorning found 60% of US employers were already using monitoring software to track remote workers, People Matters said in that review. Another 17% of employers were considering adoption, according to the same HRMorning survey as relayed by People Matters. Employers cited three objectives in that survey, People Matters said: understanding how employees spent their time, confirming employees worked a full day, and preventing personal use of company equipment.

Several employers reported productivity improvements after implementation, People Matters said. Questions emerged even in 2021 about whether those observed gains reflected genuine performance improvements or behavioural changes driven by surveillance, the publication reported. Whether the spending delivered the outcomes organisations sought remains unresolved, and the evidence suggests a more complicated picture than many early advocates anticipated, according to People Matters.

Tron's take

My read is that the 2035 GDP projection is the least usable number in this story for a US small or mid-sized business. The payback window Salesforce reported is the figure I would test, because a return-on-investment claim is one a single workflow's costs and measured savings can check inside a fiscal year.

I would keep the population straight. Salesforce described India-focused research covering business leaders in India and Global Capability Centres, so I would not read the customer satisfaction result as a forecast for a 20-person shop in Chicago. That is my reading of the news, not a reported result.

The monitoring history People Matters published is the caution I would hold next to any vendor payback number. Employers bought those tools on promised productivity gains and are still arguing five years later about whether the gains were real. My advice is to take a baseline measurement before agents go in, not after, so the comparison exists when someone asks.

Practically: pick one process you already measure, run agents against it, and compare what it cost against what it saved over the window Salesforce published. Because Salesforce described agents deployed across customer service, operations and business workflows, those agents will need credentials, logging and a rollback path before they touch live systems. XL.net sells managed IT and security assessments, so that is work we do and I am naming the conflict plainly. Ignoring the release entirely is the other error; knowing what shipped is how adoption timing stays deliberate.

For related coverage, see OpenAI Weighs AI Slowdown Law: Adoption Timing and AI Models Move Faster Than Rules Firms Track.