In the self-proclaimed year of artificial intelligence (AI) agents, real-world examples have been relatively scarce despite mountains of press releases and megaphone declarations from tech executives.
But they do exist and have proven effective in their initial deployments, based on more than a dozen interviews with AI agent vendors and their customers in healthcare, finance, retail, logistics, and manufacturing—fields that are increasingly adopting autonomous agents to do everything from automating schedules and workflows to accelerating drug discovery.
“AI adoption has accelerated significantly this year, with healthcare organizations moving beyond experimentation to deploy real-world solutions that deliver measurable value,” Jesse Cugliotta, global head of healthcare and life sciences at Snowflake, said in an email. He noted AI is transitioning from back-office applications to human-centered solutions that directly impact front-line workflows.
Market estimates widely vary in the size of the current AI agent market, with a consensus of between $5 billion and $7 billion. By 2030, analysts are forecasting between $50 billion and $100 billion.
“We are seeing an agentification of sorts of people and businesses gaining access to superpowers such as AI coding and taking notes,” said Deon Nicholas, president and co-founder of Forethought, which builds customer service AI agents for small- to mid-sized companies. “We are seeing it adopted in ways that are not hype.”
Nicholas points to Forethought’s annual report on AI use that found companies are “applying AI at higher rates in most use cases, and they’re more consistently choosing optimal approaches to do so: agentic vs. RAG-only, dedicated solutions vs. add-ons or in-house build, historic company data vs. public data.”
For months, there’s been a significant gap between AI agent announcements from nearly every major tech company and actual implementation use cases at organizations, where agents are being tested and assessed. The delay has been attributed to choosing from an overwhelming array of options, upskilling employees, and implementing governance plans.
The impact for now, however, is negligible. Nearly eight in 10 companies have reported using generative AI, but just as many have reported “no significant bottom-line impact,” says recent research from McKinsey & Company.
“Every single day there are announcements. How do you make sense of it all?” said Annie Shea Weckesser, chief marketing officer at SambaNova.ai. “Customers are trying to figure out their use cases, specific to sales agents and marketing and SDR cases. The same thing happened with generative AI. It felt like everyone was pivoting strategy.”
OpenAI’s introduction last month of an AI agent in ChatGPT—capable of helping users run code, navigate personal calendars, and generate presentations and slideshows—has heightened enterprise interest in agent capabilities.
“There is traction in restaurant and travel booking fields, especially since OpenAI’s ChatGPT agent mode has made it possible to not only research vacations but proceed to book trips autonomously,” said Peter Horadan, CEO of Vouched, a company behind technology that performs identity verification for AI agents.
Several fields have emerged as leaders in the use of AI agents:
Healthcare
Gilead Sciences has partnered with Cognizant to build and deploy a multi-agent AI system to oversee its IT operations and enable key business functions like finance and HR to communicate through an interconnected system of LLM-powered agents.
The new system tackles IT operations end-to-end—from laptop requests and new software installations to onboarding and offboarding employees, contractors, and suppliers. IT processes that previously took weeks and involved countless human and technology touchpoints are, in early testing, being reduced to a couple of days and navigated through a single AI agent that manages various workflows staffed by other agents.
Biopharmaceutical giant Madrigal Pharmaceuticals uses AI to accelerate drug discovery, optimize clinical operations, and enhance patient engagement. It operates more than 50 AI solutions, including agents, internal GPTs, and custom ML models. With a small AI governance council and limited formal processes, the organization faced growing pains in managing AI responsibly, so it partnered with Credo AI to design and implement a comprehensive AI governance program tailored to pharma-specific risks.
AstraZeneca used Databricks’ Agent Bricks to internally parse more than 400,000 clinical trial documents and extract structured data points without writing a single line of code. In roughly an hour, company officials said they created an agent that can transform complex unstructured data into usable analytics.
Still, AI agents remain rare in U.S. healthcare, says Dr. Matt Crowson, director of AI and Generative AI product management at Wolters Kluwer. “Most AI today assists clinicians under tight supervision, but a handful of systems already operate largely autonomously inside hospitals,” he said.
Finance
Brazil’s Bradesco bank used IBM Corp.’s Watson to assist employees with internal customer support and later expanded to direct customer-facing channels. Through training in Portuguese and banking context, Watson achieved a 95% accuracy rate across more than 10 million interactions, reducing response times from 10 minutes to a few seconds. It now handles roughly 283,000 inquiries monthly with minimal escalation.
A six-week pilot at ANZ Bank involving 1,000 engineers produced productivity gains, though code security outcomes remained inconclusive under a GitHub Copilot use case.
“Small business lending is a perfect real-world example where agentic AI can thrive. Traditionally underserved by digital innovation, this segment is still weighed down by legacy systems and in-person processes,” said Kos Joshi, chief business officer at iBusiness Funding, maker of a multi-agent lending tool for financial institutions.
Manufacturing
Salesforce Inc. is working with several manufacturers, including Regal Rexnord, which uses AI to support service agents in creating personalized customer responses, automatically pulling data from ERP and order management systems. Another customer, Panasonic, deploys AI agents alongside humans to triage customer requests and provide accurate answers. Nearly three-quarters of agent conversations are resolved without human intervention.
Supply-chain software company Kinaxis recently launched an AI agents program under the Maestro platform, already in use by more than a dozen global manufacturers and life sciences companies. Kinaxis’ “digital teammates” work within live supply chain environments to monitor real-time data across suppliers, plants, and logistics partners; surface recommendations when conditions shift; and help execute chosen plans, cutting decision cycles from hours to minutes.
Retail
Tredence said it is working with the world’s largest manufacturer of confectionery and pet food, where agentic AI reads free-text clinical notes from veterinarians and generates recommendations for diagnosis or treatment. This approach has cut decision-making time in half, improved pet health outcomes, and supported veterinary staff with intelligent copilots.
A collaboration between Parloa, Genesys, and Future of Voice allowed major German sporting goods retailer Decathlon to roll out AI-based service bots across phone, chat, and messenger channels. Using conversational AI, the company has optimized the customer experience across 80 locations.
The Reality Check
If there was any doubt 2025 was the year of agentic AI, those fleeting thoughts were dispelled by an avalanche of surveys, polls, anecdotes, and breathless commentary from the very companies supplying agentic AI.
It’s just not happening as quickly or extensively as everyone expected.
Vernon Keenan, an AI expert who works closely with Salesforce Inc., says agentic AI has struggled to gain traction with “organic uses” at enterprises due to data readiness issues, citing recent studies from Gartner and RAND. He doesn’t expect substantial agentic AI adoption to materialize until early 2026.
The adoption of AI agents, which are expected to take off over the next year to 18 months, may delight cost-cutting corporate executives eyeing the bottom line, but it is sending a chill through the ranks of workers—many of whom are unfamiliar with or wary of AI.
Microsoft CEO Satya Nadella and Amazon CEO Andy Jassy have only intensified fears by openly embracing AI and boasting of productivity gains while slashing workforces.
The proliferation of AI programming tools, coupled with tens of thousands of layoffs at Microsoft Corp., Meta Platforms Inc., Intel Corp., and Amazon.com Inc., has diminished job prospects for many—especially recent college graduates majoring in computer engineering and computer science.
Unemployment rates for those groups are 7.5% and 6.1%, respectively, according to a report from the Federal Reserve Bank of New York. The rates are double the unemployment rate of recent biology and art history graduates, which hover at 3%.
Despite bold claims from tech giants, most AI agents will fail when deployed in real-world enterprise settings. Current AI systems struggle with data readiness, security, and operational complexity, making them ill-equipped for mission-critical business functions in industries like retail, finance, IT, and manufacturing.
The biggest hurdles remains data readiness. Healthcare data, for example, is becoming increasingly complex, with growing volumes of multimodal data existing in formats that aren’t naturally interoperable. Many organizations have extensive data assets but struggle with fragmented silos, Snowflake’s Cugliotta said.
“This reinforces a fundamental truth: AI is only as effective as the data it’s built on and a unified data foundation is the essential foundation for AI success,” Cugliotta said.
Originally published on Techstrong. For more details, visit the source.