{"id":4233,"date":"2026-09-01T09:00:00","date_gmt":"2026-09-01T09:00:00","guid":{"rendered":"https:\/\/palegoldenrod-boar-492303.hostingersite.com\/?p=4233"},"modified":"2026-08-30T13:11:06","modified_gmt":"2026-08-30T13:11:06","slug":"agentic-ai-examples","status":"publish","type":"post","link":"https:\/\/growthnation.ai\/insights\/agentic-ai-examples\/","title":{"rendered":"Agentic AI Examples: What It Looks Like in Practice"},"content":{"rendered":"<span class=\"span-reading-time rt-reading-time\" style=\"display: block;\"><span class=\"rt-label rt-prefix\">Reading Time: <\/span> <span class=\"rt-time\"> 13<\/span> <span class=\"rt-label rt-postfix\">minutes<\/span><\/span><div class=\"\">\n\t\t\t<div class=\"aioseo-toc-header\">\n\t\t\t\t<header class=\"aioseo-toc-header-area\">\n\t\t\t\t\t<div class=\"aioseo-toc-header-title aioseo-toc-header-collapsible-closed \">\n\t\t\t\t\t<div class=\"aioseo-toc-header-collapsible\">\n\t\t\t\t\t\t<svg width=\"14\" height=\"14\" viewBox=\"0 0 14 14\" fill=\"none\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\">\n\t\t  <path d=\"M6 8H0V6H6V0H8V6H14V8H8V14H6V8Z\" fill=\"#005AE0\"\/>\n\t\t<\/svg>\n\t\t\t\t\t<\/div>\n\t\t\t\t\tShow Table of Contents\n\t\t\t\t\t<\/div>\n\n\t\t\t\t\t<div class=\"aioseo-toc-header-title aioseo-toc-header-collapsible-open aioseo-toc-collapsed\">\n\t\t\t\t\t<div class=\"aioseo-toc-header-collapsible\">\n\t\t\t\t\t\t<svg width=\"14\" height=\"2\" viewBox=\"0 0 14 2\" fill=\"none\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\">\n\t\t  <path d=\"M0 2V0H14V2H0Z\" fill=\"#005AE0\"\/>\n\t\t<\/svg>\n\t\t\t\t\t<\/div>\n\t\t\t\t\tHide Table of Contents\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/header>\n\t\t\t\t<div class=\"aioseo-toc-contents aioseo-toc-collapsed\">\n\t\t\t\t\t<ul><li><a class=\"aioseo-toc-item\" href=\"#aioseo-what-counts-as-an-agentic-ai-example-4\">What counts as an agentic AI example<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-agentic-ai-examples-by-business-function-11\">Agentic AI examples by business function<\/a><ul><li><a class=\"aioseo-toc-item\" href=\"#aioseo-customer-service-at-salesforce-13\">Customer service at Salesforce<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-software-engineering-at-google-18\">Software engineering at Google<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-logistics-at-c-h-robinson-23\">Logistics at C.H. Robinson<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-field-operations-at-openreach-27\">Field operations at Openreach<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-banking-at-lloyds-banking-group-31\">Banking at Lloyds Banking Group<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-public-services-at-citizens-advice-36\">Public services at Citizens Advice<\/a><\/li><\/ul><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-what-the-working-examples-have-in-common-41\">What the working examples have in common<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-where-agentic-ai-examples-fall-short-51\">Where agentic AI examples fall short<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-how-to-judge-an-example-before-you-copy-it-56\">How to judge an example before you copy it<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-frequently-asked-questions-67\">Frequently asked questions<\/a><ul><li><a class=\"aioseo-toc-item\" href=\"#aioseo-what-is-an-example-of-agentic-ai-68\">What is an example of agentic AI?<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-what-is-the-difference-between-an-ai-agent-and-agentic-ai-70\">What is the difference between an AI agent and agentic AI?<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-are-ai-agents-actually-being-used-in-production-72\">Are AI agents actually being used in production?<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-which-business-functions-use-agentic-ai-most-74\">Which business functions use agentic AI most?<\/a><\/li><li><a class=\"aioseo-toc-item\" href=\"#aioseo-how-can-you-tell-if-something-is-really-an-agent-76\">How can you tell if something is really an agent?<\/a><\/li><\/ul><\/li><\/ul>\n\t\t\t\t<\/div>\n\t\t\t<\/div>\n\t\t<\/div>\n\n\n<p>Stanford HAI reported that organisational AI adoption reached 88 per cent in 2025, up from 78 per cent the year before. Agent deployment tells a different story. Across nearly every business function, scaled agent use sits in the single digits, and in most functions a majority of respondents report no agent use at all. Even in IT and knowledge management, the two busiest areas, around two thirds or more report none. The technology sector leads, at 24 per cent scaled use in software engineering, 22 per cent in IT and 21 per cent in service operations.<\/p>\n\n\n\n<p>So the useful question is not whether agents are coming. It is what the small minority already running them is actually doing.<\/p>\n\n\n\n<p><em><strong>Quick Answer: <\/strong>Agentic AI examples running in production today include Salesforce resolving most inquiries on its own support site without a human, Google using a three-agent system to migrate production code, C.H. Robinson passing shipment data from one agent to another, and Openreach contacting customers to rebook engineer appointments. Genuine examples share a pattern. The system is given a goal rather than a prompt, works across several steps, uses real tools, checks its own output, and involves a person only when something falls outside the expected path.<\/em><\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"aioseo-what-counts-as-an-agentic-ai-example-4\"><strong>What counts as an agentic AI example<\/strong><\/h2>\n\n\n\n<p>Most published lists of agentic AI examples describe a hypothetical. A support agent receives a ticket, checks a customer history, drafts a reply. Nobody is named, nothing is dated, and no figure can be checked. That makes for a readable page and a weak basis for a decision.<\/p>\n\n\n\n<p>Gartner introduced the term agent washing in June 2025 to describe the rebranding of assistants, scripted automation and chatbots as agentic products. In the same analysis it estimated that only around 130 of the thousands of vendors claiming agentic capability were building the real thing. The date matters, because the figure has circulated widely since without it.<\/p>\n\n\n\n<p>Gartner also offers a practical test. Use an agent when a decision is required, automation when the workflow is routine, and an assistant when the need is simple retrieval. Many use cases described as agentic do not require an agent at all.<\/p>\n\n\n\n<p>MIT Sloan draws a further distinction worth borrowing. Sinan Aral describes a single AI agent as something that acts, while agentic AI more precisely covers several agents orchestrating a task together. The underlying concepts are covered separately in <a href=\"https:\/\/growthnation.ai\/insights\/what-is-an-ai-agent\/\">what an AI agent is<\/a> and <a href=\"https:\/\/growthnation.ai\/insights\/what-is-agentic-ai\/\">agentic AI<\/a> separately, along with the difference between <a href=\"https:\/\/growthnation.ai\/insights\/agentic-ai-vs-generative-ai\/\">generative and agentic AI<\/a>.<\/p>\n\n\n\n<p>A verifiable example clears one bar: a named organisation, a source published by that organisation or its named partner, a date, and a reported figure. Anything short of that is an illustration rather than evidence.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><td><strong>Organisation<\/strong><\/td><td><strong>Function<\/strong><\/td><td><strong>What the agent does<\/strong><\/td><td><strong>Reported result<\/strong><\/td><td><strong>Source and date<\/strong><\/td><\/tr><\/thead><tbody><tr><td>Salesforce<\/td><td>Customer service<\/td><td>Answers questions and manages cases across web, portal, voice and messaging on its own support site<\/td><td>4.3 million inquiries handled, 70 per cent resolved without a human<\/td><td>Salesforce newsroom, 25 June 2026<\/td><\/tr><tr><td>Google<\/td><td>Software engineering<\/td><td>Three specialist agents plan, sequence and rewrite production model code, testing and fixing until it builds<\/td><td>6.4 to 8 times faster than manual migration on live YouTube models<\/td><td>Google Cloud blog, 6 May 2026<\/td><\/tr><tr><td>C.H. Robinson<\/td><td>Logistics<\/td><td>One agent extracts shipment updates from phone calls and passes them to a second agent that writes them into the platform<\/td><td>318,000 tracking updates captured from one call type in a single month<\/td><td>Company announcement, 20 October 2025<\/td><\/tr><tr><td>Openreach<\/td><td>Field operations<\/td><td>Starts contact with customers by text, email and voice to confirm, move or rebook engineer appointments<\/td><td>Missed appointments and inbound contacts each down by about a third<\/td><td>NiCE press release, 7 April 2026<\/td><\/tr><tr><td>Lloyds Banking Group<\/td><td>Banking<\/td><td>Assistive tools in production across search, coding and HR queries, with agentic use cases scheduled for 2026<\/td><td>Around 50 million pounds of value from generative AI in 2025<\/td><td>Lloyds press release, 29 January 2026<\/td><\/tr><tr><td>Citizens Advice<\/td><td>Public services<\/td><td>Drafts answers from approved sources for an adviser to check before the client sees them<\/td><td>Described by its builders as a copilot, with human review required<\/td><td>i.AI project pages, 2024 to 2026<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"aioseo-agentic-ai-examples-by-business-function-11\"><strong>Agentic AI examples by business function<\/strong><\/h2>\n\n\n\n<p>Agent deployment is not spread evenly. It clusters in functions with a high volume of repeatable work and an output that can be checked, which is why customer service, software engineering and operations account for most documented activity.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"aioseo-customer-service-at-salesforce-13\"><strong>Customer service at Salesforce<\/strong><\/h3>\n\n\n\n<p>Salesforce runs an agent on its own support site. Announcing a packaged version of it in June 2026, the company reported that the agent had handled 4.3 million inquiries and resolved 70 per cent of them without a person stepping in.<\/p>\n\n\n\n<p>The agent does more than answer questions. It manages cases and runs actions such as scheduling appointments and updating orders, across web, portal, voice and messaging channels.<\/p>\n\n\n\n<p>Two caveats belong with that figure. It is a vendor reporting on its own product, so treat it as a supplier claim rather than independent evidence. Salesforce also cites a resolution rate above 75 per cent for the same deployment elsewhere on its site, so the number varies by source.<\/p>\n\n\n\n<p>The pricing model is itself a claim about autonomy. Customers are charged per resolution, with no charge when someone escalates to a human or abandons the conversation.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"aioseo-software-engineering-at-google-18\"><strong>Software engineering at Google<\/strong><\/h3>\n\n\n\n<p>Google built a multi-agent system to migrate production machine learning models from one framework to another, published on the Google Cloud blog in May 2026 alongside a technical paper.<\/p>\n\n\n\n<p>Three specialist agents divide the job. A planner maps the dependency tree using compiler-based static analysis and breaks the migration into ordered steps. An orchestrator groups those steps, supplies the necessary domain knowledge and handles recovery when a step fails to build. A coder reads files, writes code, runs builds and executes tests, working in a test and fix loop until each component compiles and passes.<\/p>\n\n\n\n<p>On real YouTube models containing thousands of lines of code and hundreds of layers, the system ran 6.4 to 8 times faster than manual migration. Engineers review the output rather than translating by hand.<\/p>\n\n\n\n<p>The most instructive detail is the failure that came first. Google reports that generic single-agent setups struggled at this scale, overwriting critical files and skipping functionality. Specialisation and verification were what made the difference.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"aioseo-logistics-at-c-h-robinson-23\"><strong>Logistics at C.H. Robinson<\/strong><\/h3>\n\n\n\n<p>The logistics provider C.H. Robinson operates a connected group of more than 30 AI agents across the shipment lifecycle, set out in its October 2025 announcement.<\/p>\n\n\n\n<p>The clearest illustration is a handoff between two of them. The company reported that in a single month, one agent captured 318,000 freight tracking updates from one type of phone call. That information had previously been invisible to its systems. A second agent then wrote those updates into the platform, where they feed predicted arrival times for customers.<\/p>\n\n\n\n<p>Neither agent completes the task alone. One pulls meaning out of unstructured speech, the other acts on it. This is close to what Aral means by agentic AI rather than a single agent.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"aioseo-field-operations-at-openreach-27\"><strong>Field operations at Openreach<\/strong><\/h3>\n\n\n\n<p>Openreach, the BT Group network business, deployed proactive agents across 15 million customer journeys during its full fibre rollout, reported by its technology partner in April 2026.<\/p>\n\n\n\n<p>The word proactive is doing real work here. Rather than waiting to be contacted, the agents start conversations by text, email and voice when something changes. They explain the upgrade, answer questions, and book, move or bring forward engineer appointments on the customer\u2019s behalf.<\/p>\n\n\n\n<p>Openreach reported that missed appointments and inbound contact volumes each fell by around a third, and that its Trustpilot rating rose from 2.0 to 4.7 across hundreds of thousands of reviews. Its director of customer service described tens of millions of pounds in benefits for Openreach and the providers it serves.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"aioseo-banking-at-lloyds-banking-group-31\"><strong>Banking at Lloyds Banking Group<\/strong><\/h3>\n\n\n\n<p>Lloyds Banking Group is among the most public UK banks on AI, and its January 2026 results disclosure is unusually specific about what is already running and what is still planned.<\/p>\n\n\n\n<p>The named, quantified tools in production are assistive rather than agentic. Athena, an internal knowledge assistant, is used by 20,000 colleagues and has cut search times by 66 per cent on average. Around 5,000 engineers use AI coding tools. An HR assistant resolves roughly 90 per cent of colleague queries correctly on first contact. Generative AI delivered around 50 million pounds of value in 2025 across more than 50 use cases.<\/p>\n\n\n\n<p>Agentic AI appears in the same disclosure as the 2026 programme, with more than 100 million pounds of additional value expected and an AI academy planned for 67,000 colleagues.<\/p>\n\n\n\n<p>That sequence is the Stanford finding visible in one company\u2019s accounts. At an organisation leading its sector on AI, the proven work is assistive and the agentic work is scheduled. That gap is the difference between <a href=\"https:\/\/growthnation.ai\/insights\/agentic-ai-vs-generative-ai\/\">generative and agentic AI<\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\" id=\"aioseo-public-services-at-citizens-advice-36\"><strong>Public services at Citizens Advice<\/strong><\/h3>\n\n\n\n<p>Caddy is an AI assistant built by Citizens Advice SORT with the UK government\u2019s AI incubator, supporting advisers handling questions on benefits, debt, housing and tax.<\/p>\n\n\n\n<p>It drafts answers from a defined set of trusted sources, including GOV.UK and internal guidance, shows where each answer came from, and routes responses for human checking before they reach the client.<\/p>\n\n\n\n<p>Its builders describe it as a copilot for frontline contact centres rather than an agent, and that description is accurate. Retrieval with a person in the loop is useful, and it is not a system acting on a goal.<\/p>\n\n\n\n<p>This is the category most often relabelled as agentic. On the Gartner test it is the assistant case, and an assistant is the right tool for the job.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"aioseo-what-the-working-examples-have-in-common-41\"><strong>What the working examples have in common<\/strong><\/h2>\n\n\n\n<p>Across the deployments that are working, the same shape recurs.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>A goal rather than a prompt. Each system is pointed at an outcome and works out the steps itself.<\/li>\n\n\n\n<li>Real tool use. Each reads from and writes to production systems rather than producing text for someone to paste elsewhere.<\/li>\n\n\n\n<li>A checking loop. Google tests and fixes until code builds. Salesforce counts a resolution only when nothing escalates.<\/li>\n\n\n\n<li>Handoff between specialists. The strongest examples pass work between narrow agents rather than asking one agent to do everything.<\/li>\n\n\n\n<li>People at the exceptions. Humans review, approve or take over at the edges, not at every step.<\/li>\n<\/ul>\n\n\n\n<p>Narrow scope is the other common factor. None of these agents was asked to run a department. Each was aimed at one repeatable, high volume task with an output that can be checked.<\/p>\n\n\n\n<p>The effort also lands somewhere unexpected. MIT Sloan researchers deploying an agent to detect adverse events in clinical notes found roughly 80 per cent of the work went on data engineering, stakeholder alignment, governance and workflow integration, rather than prompting or model tuning. The hard part was rarely the AI itself. There is more on this in <a href=\"https:\/\/growthnation.ai\/insights\/ai-agents-for-business\/\">AI agents in business<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"aioseo-where-agentic-ai-examples-fall-short-51\"><strong>Where agentic AI examples fall short<\/strong><\/h2>\n\n\n\n<p>The picture is not uniformly positive.<\/p>\n\n\n\n<p>In June 2025 Gartner forecast that more than 40 per cent of agentic AI projects would be cancelled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls. The stated causes are scoping and governance problems rather than limits on what the models can do.<\/p>\n\n\n\n<p>Operational risk is measurable too. The IBM Institute for Business Value surveyed 2,000 technology executives in early 2026 and found an average of 54 AI agent incidents per organisation over the previous year, meaning unintended or harmful occurrences requiring human correction. Around 17 per cent were high severity, taking more than four hours to contain. In the same study, 77 per cent said AI adoption was outpacing their governance, and only 11 per cent felt ready for the scale of deployment expected within the year.<\/p>\n\n\n\n<p>Measurement deserves the same care. As MIT Sloan\u2019s Kate Kellogg puts it, reclaiming a fifth of someone\u2019s time is not the same as saving a fifth of the labour cost. If you are weighing up your own position before building anything, an <a href=\"https:\/\/growthnation.ai\/insights\/ai-readiness-assessment\/\">AI readiness assessment<\/a> and an <a href=\"https:\/\/growthnation.ai\/insights\/ai-maturity-assessment\/\">AI maturity assessment<\/a> are the places to start.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"aioseo-how-to-judge-an-example-before-you-copy-it-56\"><strong>How to judge an example before you copy it<\/strong><\/h2>\n\n\n\n<p>Some published examples can be checked and most cannot. Five questions separate the two.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Is a specific organisation named, or only a role and an industry?<\/li>\n\n\n\n<li>Is the figure traceable to a dated, published source?<\/li>\n\n\n\n<li>Does the outcome describe activity, such as tasks handled, or resolution, such as work finished without a person?<\/li>\n\n\n\n<li>Would the task genuinely need an agent, or would automation or retrieval do it?<\/li>\n\n\n\n<li>Is someone accountable when the agent gets it wrong?<\/li>\n<\/ul>\n\n\n\n<p>The same questions apply internally. Organisations that keep a central record of the agents they own, with hours saved recorded against each one, can answer them about their own deployments. Tracking that position over time is what an <a href=\"https:\/\/growthnation.ai\/insights\/ai-scorecard\/\">AI scorecard<\/a> and an <a href=\"https:\/\/growthnation.ai\/insights\/ai-maturity-model\/\">AI maturity model<\/a> are for.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><td><strong>Source<\/strong><\/td><td><strong>What it covers<\/strong><\/td><td><strong>Useful for<\/strong><\/td><td><strong>Published<\/strong><\/td><\/tr><\/thead><tbody><tr><td>Stanford HAI AI Index, Chapter 4<\/td><td>Organisational AI adoption and agent deployment rates by business function<\/td><td>Judging how common agents actually are<\/td><td>2026<\/td><\/tr><tr><td>MIT AI Agent Index<\/td><td>Public database from MIT CSAIL documenting agentic systems in use<\/td><td>Finding further named deployments<\/td><td>Ongoing<\/td><\/tr><tr><td>Gartner newsroom release on agentic AI<\/td><td>Forecast cancellation rate, and the difference between agents, automation and assistants<\/td><td>Testing whether a use case needs an agent<\/td><td>25 June 2025<\/td><\/tr><tr><td>IBM Institute for Business Value study<\/td><td>Survey of 2,000 technology executives on agent scale, governance and incidents<\/td><td>Understanding operational risk<\/td><td>8 June 2026<\/td><\/tr><tr><td>MIT Sloan, Agentic AI explained<\/td><td>Academic view on definitions, implementation effort and measurement<\/td><td>Setting expectations before you build<\/td><td>18 February 2026<\/td><\/tr><tr><td>GrowthNation Agent Vault and analytics<\/td><td>Central record of the agents an organisation owns and the hours they save<\/td><td>Tracking your own deployments<\/td><td>Ongoing<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>If you would rather see this on your own work than on someone else\u2019s, we map a single team, surface the repeatable tasks worth automating, and show you the agent that comes out of it. <a href=\"https:\/\/growthnation.ai\/#how-it-works\">Start with one interview<\/a>.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\" id=\"aioseo-frequently-asked-questions-67\"><strong>Frequently asked questions<\/strong><\/h2>\n\n\n\n<div data-schema-only=\"false\" class=\"wp-block-aioseo-faq\" id=\"aioseo-what-is-an-example-of-agentic-ai-68\"><h3 class=\"aioseo-faq-block-question\"><strong>What is an example of agentic AI?<\/strong><\/h3><div class=\"aioseo-faq-block-answer\">\n<p>Google\u2019s model migration system is among the clearest. Three specialist agents plan the work, sequence it and rewrite production code, testing and fixing until it builds. Google reports the process running 6.4 to 8 times faster than migrating the same models by hand.<\/p>\n<\/div><\/div>\n\n\n\n<div data-schema-only=\"false\" class=\"wp-block-aioseo-faq\" id=\"aioseo-what-is-the-difference-between-an-ai-agent-and-agentic-ai-70\"><h3 class=\"aioseo-faq-block-question\"><strong>What is the difference between an AI agent and agentic AI?<\/strong><\/h3><div class=\"aioseo-faq-block-answer\">\n<p>The terms are often used interchangeably. The more precise distinction is that <a href=\"https:\/\/growthnation.ai\/insights\/what-is-an-ai-agent\/\">a single AI agent<\/a> acts on a goal, while agentic AI describes several agents orchestrating a task together, as in the C.H. Robinson handoff between two agents.<\/p>\n<\/div><\/div>\n\n\n\n<div data-schema-only=\"false\" class=\"wp-block-aioseo-faq\" id=\"aioseo-are-ai-agents-actually-being-used-in-production-72\"><h3 class=\"aioseo-faq-block-question\"><strong>Are AI agents actually being used in production?<\/strong><\/h3><div class=\"aioseo-faq-block-answer\">\n<p>Yes, but far less widely than coverage suggests. Stanford HAI reports 88 per cent organisational AI adoption alongside scaled agent use in the single digits across nearly every business function, with a majority of respondents reporting no agent use at all.<\/p>\n<\/div><\/div>\n\n\n\n<div data-schema-only=\"false\" class=\"wp-block-aioseo-faq\" id=\"aioseo-which-business-functions-use-agentic-ai-most-74\"><h3 class=\"aioseo-faq-block-question\"><strong>Which business functions use agentic AI most?<\/strong><\/h3><div class=\"aioseo-faq-block-answer\">\n<p>Software engineering, IT and service operations, and mainly within the technology sector, where scaled use reaches 24, 22 and 21 per cent respectively. Uptake is markedly lower in manufacturing, supply chain and human resources.<\/p>\n<\/div><\/div>\n\n\n\n<div data-schema-only=\"false\" class=\"wp-block-aioseo-faq\" id=\"aioseo-how-can-you-tell-if-something-is-really-an-agent-76\"><h3 class=\"aioseo-faq-block-question\"><strong>How can you tell if something is really an agent?<\/strong><\/h3><div class=\"aioseo-faq-block-answer\">\n<p>Ask whether it pursues a goal across multiple steps without a fresh prompt each time, whether it acts on real systems rather than only producing text, and whether it checks its own work. If a person has to approve every step, it is an assistant.<\/p>\n<\/div><\/div>\n\n\n\n<p><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Stanford HAI reported that organisational AI adoption reached 88 per cent in 2025, up from 78 per cent the year before. Agent deployment tells a different story. Across nearly every business function, scaled agent use sits in the single digits, and in most functions a majority of respondents report no agent use at all. Even [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":4234,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"content-type":"","_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[16],"tags":[],"class_list":["post-4233","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-thoughts"],"aioseo_notices":[],"jetpack_featured_media_url":"https:\/\/i0.wp.com\/growthnation.ai\/insights\/wp-content\/uploads\/2026\/08\/agentic-ai-examples.webp?fit=1672%2C941&ssl=1","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/posts\/4233","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/comments?post=4233"}],"version-history":[{"count":1,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/posts\/4233\/revisions"}],"predecessor-version":[{"id":4235,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/posts\/4233\/revisions\/4235"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/media\/4234"}],"wp:attachment":[{"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/media?parent=4233"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/categories?post=4233"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/growthnation.ai\/insights\/wp-json\/wp\/v2\/tags?post=4233"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}