AI Agents Trends 2027
By Axis Intelligence Research
Co-author: Sarah Mitchell | Last updated: October 28, 2026 | License: CC BY 4.0
Axis Intelligence Research projects worldwide spending on AI agents and assistants of about $44 billion in 2027 in its base case. That is roughly a third below Gartner’s September 2026 forecast of $65.5 billion — the only published 2027 figure for this scope — and the gap comes down to one question: how many large enterprises actually scale agents next year.
Quick Answer
Axis Intelligence Research’s base case puts 2027 spending on AI agents and assistants at about $44 billion, inside a $35–54 billion scenario range and below Gartner’s $65.5 billion forecast (September 16, 2026). The driver most likely to move the number is production scaling at large enterprises: 40% were scaling agents in mid-2026, per McKinsey.
Key Findings
- Axis Intelligence Research projects about $44 billion of worldwide AI agent and assistant spending in 2027 in its base case, 50% above Gartner’s 2026 estimate of $29.2 billion.
- Gartner’s $65.5 billion 2027 forecast implies that about 75% of large enterprises scale AI agents by mid-2027, versus 40% in McKinsey’s 2026 survey, according to an Axis Intelligence Research calculation.
- Published estimates of the AI agent market disagree by 2.3x on 2025, a year already closed: Fortune Business Insights puts agentic AI at $7.29 billion; Gartner puts agents and assistants at $16.5 billion.
- 40% of organizations with revenue above $1 billion reported scaling AI agents in 2026, up from 27% a year earlier, while smaller organizations held at 22%, per McKinsey (August 25, 2026).
- 32% of McKinsey’s 2026 respondents say their organization decided against buying at least one software product because agentic coding tools let them build it in-house.
How Much Will Companies Spend on AI Agents in 2027?
Axis Intelligence Research projects about $44 billion in 2027 in its base case, measured in the same scope Gartner uses for its “AI agents and assistants” spending segment. Gartner’s own figure for 2027 is $65.5 billion, published September 16, 2026, from a 2025 base of $16.5 billion. Gartner estimates 2026 at $29.2 billion.
Every figure in this section is a projection or an estimate. Gartner’s 2026 number is itself a forecast for a year that has two months left to run; it is the starting point of the Axis model, not a reported outcome.
The Axis Base Case: how the number is built
Most agent market forecasts apply a single growth rate to a market total. The Axis Base Case splits 2027 spending into two drivers that can be checked separately — how many buyers scale agents, and how much each of them spends:
Spending 2027 = Spending 2026 × adopter growth × spend-per-adopter growth
- Spending 2026 = $29.219 billion. Gartner’s estimate for AI agents and assistants, September 16, 2026.
- Adopter growth. The share of large organizations (annual revenue above $1 billion) scaling AI agents: 27% in McKinsey’s 2025 survey, 40% in its 2026 survey. The base case assumes a further 10-point rise to 50% by mid-2027 — slower than the 13-point jump of the past year, because Gartner expects more than 40% of agentic AI projects to be canceled by the end of 2027. Factor: 50 ÷ 40 = 1.25.
- Spend-per-adopter growth. Backed out from the past year. Gartner’s segment grew 1.773x from 2025 to 2026; large-organization adopters grew 1.481x (40 ÷ 27). The remainder — 1.773 ÷ 1.481 ≈ 1.20 — is the implied growth in spend per adopting organization. The base case holds it at 1.20.
$29.219B × 1.25 × 1.20 = $43.8 billion, rounded to about $44 billion because the adopter assumption cannot justify more precision than that.
Adopter growth is measured on large organizations only. That is an assumption, not a sourced fact: Axis treats large enterprises as the marginal buyer because McKinsey’s data shows smaller organizations flat at 22% scaling in both years. If small-company adoption breaks out in 2027, this model understates spending.
Base, upside and downside scenarios for 2027
| Scenario | 2027 spending (Axis projection) | Large-org scaling share, mid-2027 | Spend per adopter | What separates it |
|---|---|---|---|---|
| Downside | about $35 billion | 45% (factor 1.125) | +5% (1.05) | Cancellations bite; falling token prices pass straight through to buyers |
| Base | about $44 billion | 50% (1.25) | +20% (1.20) | Scaling slows from last year’s pace; per-adopter spend grows at its 2025–26 rate |
| Upside | about $54 billion | 53% (1.325) | +40% (1.40) | Last year’s 13-point scaling jump repeats; usage-based agent pricing lifts wallets |
| Gartner (for reference) | $65.5 billion | — | — | Gartner forecast, September 16, 2026 |
Source: Axis Intelligence Research calculation using Gartner (September 16, 2026) and McKinsey (August 25, 2026). Scenario values: $34.5B, $43.8B, $54.2B before rounding.
Which assumption matters most?
The adopter assumption carries slightly more weight than the wallet assumption. Moving the 2027 large-organization scaling share 5 points either way (45% or 55%) shifts the base case to $39.4–48.2 billion. Moving spend-per-adopter growth 10 points either way (1.10 or 1.30) shifts it to $40.2–47.5 billion.
Here is where Axis parts company with Gartner. Holding spend per adopter at 1.20, reaching $65.5 billion would require 74.7% of large organizations scaling agents by mid-2027. Holding adoption at the base-case 50%, it would require spend per adopter to rise 79%. Neither looks consistent with a technology Gartner placed at the Peak of Inflated Expectations in its first Hype Cycle for Agentic AI (April 2, 2026).
The likelier explanation is scope. Gartner says its September forecast separated cross-functional agents and assistants from AI software and added consumer agents and assistants to the segment. A paid consumer-agent market arriving in 2027 is a step change the Axis model — built on enterprise adoption — does not capture by design. Read the gap as “enterprise demand alone gets you to the mid-$40 billions; the rest of Gartner’s number is a bet on consumers paying for agents.”
AI Agent Market Size Forecasts Compared (2025–2034)
Nine published forecasts from seven firms were reviewed. They do not measure the same thing, and the table below keeps them apart instead of averaging them.
| Firm | Published | Scope | Base year value | Horizon value | Stated CAGR | Source |
|---|---|---|---|---|---|---|
| Gartner | Sep 16, 2026 | AI agents and assistants (enterprise cross-functional + consumer) | $16.481B (2025) | $65.472B (2027) | — (implied 99.3%, 2025–27) | Gartner |
| Gartner | Feb 25, 2026 | Enterprise cross-functional AI agents and assistants | not stated | more than $23B (2030) | 59.0% | Gartner research abstract |
| MarketsandMarkets | Apr 23, 2025 | AI agents market | $7.84B (2025) | $52.62B (2030) | 46.3% (2025–30) | MarketsandMarkets |
| BCC Research | Oct 2025 | AI agents: technologies, applications, markets | $8B (2025) | $48.3B (2030) | 43.3% (2025–30) | BCC Research |
| Grand View Research | Mar 2026 | AI agents market | $7.63B (2025) | $182.97B (2033) | 49.6% (2026–33) | Grand View Research |
| Fortune Business Insights | updated Oct 12, 2026 | Agentic AI market | $7.29B (2025) | $139.19B (2034) | 40.50% (2026–34) | Fortune Business Insights |
| Precedence Research | Sep 4, 2025 | Agentic AI market | $7.55B (2025) | $199.05B (2034) | 43.84% (2025–34) | Precedence Research |
| Gartner | Aug 26, 2025 | Agentic AI share of enterprise application software revenue (best case) | 2% (2025) | ~30%, more than $450B (2035) | — | Gartner |
| IDC | Aug 26, 2025 | Total AI IT spending, which IDC attributes mainly to agentic AI | — | $1.3T (2029) | 31.9% (2025–29) | IDC |
Implied CAGRs are Axis Intelligence Research calculations from each firm’s published start and end values. No firm’s forecast has been extended or interpolated to a year it did not publish.
Why estimates of the same market differ by more than 2x
The five standalone “AI agents” or “agentic AI” studies cluster tightly on 2025: $7.29 billion to $8 billion. Gartner’s figure for the same year is $16.481 billion. The 2.3x gap is scope, not optimism. Gartner’s segment counts assistants as well as agents, and since September 2026 it includes consumer products; the standalone studies size purpose-built agent software.
The horizons scatter far more than the bases. For 2030, the three published points run from “more than $23 billion” (Gartner, enterprise cross-functional only) to $52.62 billion (MarketsandMarkets). Past 2030, the long-dated studies run to $139–199 billion by 2034 on CAGRs of 40–44% held for nine years. A compound rate that high, sustained that long, has more to do with spreadsheet mechanics than with any observed adoption curve — nothing in today’s survey data can confirm or rule out year nine.
IDC’s $1.3 trillion is a different object entirely. It is total AI IT spending in 2029, including infrastructure, which IDC’s headline attributes to agentic growth. It is not an agent software market and should not be quoted as one.
How these forecasts have moved: a track-record check
Forecasts for this market have only gone up. MarketsandMarkets’ September 2024 edition put 2030 at $47.1 billion from a 2024 base of $5.1 billion; its April 2025 edition raised 2030 to $52.62 billion (+11.7%) and restated 2024 at $5.26 billion. Gartner’s forecast of total AI spending for 2027 moved from $3.34 trillion (January 2026) to $3.49 trillion (May 2026) to $3.64 trillion (September 2026), a 9.0% upward revision in eight months.
Upward drift is not proof of error. It does mean that a reader comparing editions sees a consensus chasing reality from below, which is exactly the setting in which a forecast built on observable drivers earns its keep.
One survey offers a cleaner test. In McKinsey’s 2025 survey, 32% of respondents expected AI to reduce their organization’s headcount over the following year; in the 2026 survey, 14% reported that it actually had. Expectations about agents have so far run well ahead of measured outcomes.
Biggest AI Agent Trends for 2027
Each trend below is backed by data already moving. Trends without present-day evidence were cut.
1. Will AI agents move from pilots to production in 2027? Mostly at large enterprises.
The claim: scaled agent deployment keeps concentrating in large organizations.
The evidence: 40% of McKinsey’s 2026 respondents from organizations above $1 billion in revenue report scaling AI agents, up from 27%; smaller organizations stayed at 22% (survey fielded May 4–June 8, 2026; 1,719 participants). A year earlier, McKinsey found 23% of all respondents scaling agents somewhere and no more than 10% scaling them in any single function. Vendor disclosures point the same way: ServiceNow says agentic deployments of ServiceNow AI increased ninefold in nine months (company-reported, July 22, 2026).
What would prove it wrong: McKinsey’s 2027 survey showing small-organization scaling well above 22%, or large-organization scaling stalling near 40%.
Who it affects: mid-market software vendors, whose customers are not yet buying at the rate the enterprise numbers suggest.
2. How will AI agents be priced? By completed work, not seats.
The claim: agent revenue is being metered on tasks finished rather than users licensed.
The evidence: Salesforce reports 7.0 billion “Agentic Work Units” delivered to date across Agentforce and Slack, with 3.2 billion in its fiscal Q2 2027 alone, up 97% quarter over quarter (company-reported, August 26, 2026). Agentforce annual recurring revenue (ARR) passed $1.5 billion, up more than 240% — but Salesforce widened the metric that quarter to include Slackbot and Headless 360, so the growth rate is not like-for-like.
What would prove it wrong: major vendors reverting to per-seat agent bundles in 2027 renewals.
Who it affects: procurement and finance teams. A per-seat budget can be forecast from headcount; a per-task budget is forecast from volume, and volume is exactly what agents are deployed to raise.
3. Will agents eat SaaS spending? They are already eating some of it.
The claim: coding agents are converting software purchases into internal builds.
The evidence: 32% of McKinsey’s 2026 respondents say their organization decided against buying at least one software product or feature because it could be built in-house with agentic coding tools. About two in ten report scaling software coding agents, rising to 31% at larger enterprises.
What would prove it wrong: net revenue retention at large application vendors holding up through 2027 renewals despite this sentiment.
Who it affects: point-solution SaaS vendors with thin feature moats. Incumbents with systems of record sit on the other side of the trade — Gartner expects AI to be sold mostly by incumbent software providers throughout 2026, which is the year it places AI in the Trough of Disillusionment (January 15, 2026).
4. Are most agentic AI projects failing? Gartner expects a large share to be canceled.
The claim: 2027 brings a shakeout of pilots and of vendors relabeling old products as agents.
The evidence: Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls, and estimates only about 130 of the thousands of self-described agentic AI vendors are real (June 25, 2025) [older data — the most recent Gartner statement on project cancellations found]. McKinsey’s 2026 survey shows the value gap persisting: 37% of respondents attribute any EBIT impact to AI, essentially unchanged from 2025.
What would prove it wrong: EBIT attribution rising clearly above 37% in McKinsey’s 2027 survey.
Who it affects: buyers running five or more agent pilots with no production owner.
5. Will governance become a separate agent budget? Yes, and the EU has set a date.
The claim: controlling agents — identity, permissions, audit, security — becomes its own spending line.
The evidence: Gartner forecasts AI cybersecurity spending of $51.3 billion in 2026 and $86.0 billion in 2027, up from $25.9 billion in 2025 (May 19, 2026). Its Hype Cycle for Agentic AI lists agentic AI governance, agentic AI security and FinOps for agentic AI as profiles in their own right. In the EU, Regulation (EU) 2026/1744, the Digital Omnibus on AI, was adopted July 8, 2026 and entered into force July 27, 2026, deferring the AI Act’s obligations for stand-alone (Annex III) high-risk systems to December 2, 2027.
What would prove it wrong: governance features bundled free into agent platforms rather than sold separately.
Who it affects: any organization deploying agents in hiring, credit or other Annex III uses in the EU.
Will AI Agents Get Cheaper to Run in 2027?
Per token, yes. Per organization, probably not — and that distinction is the cost story of 2027.
Unit prices keep collapsing. Epoch AI found the price of reaching a fixed level of model performance falling between 9x and 900x per year depending on the benchmark, with a median of 50x (March 2025) [older data — no newer Epoch release of this series found]. A 2025 preprint by MIT researchers using Epoch AI and Artificial Analysis data puts the decline for frontier models at 5x to 10x per year.
Yet bills are rising faster than unit prices fall. Agents run multi-step loops — plan, call a tool, check, retry — so a single task consumes many model calls. Gartner expects inference to overtake training in AI-optimized cloud infrastructure spending in 2026 ($23.3 billion vs. $19 billion), with inference’s share reaching 59% in 2027 (August 10, 2026). About 20% of McKinsey’s 2026 respondents say AI operating costs, including tokens, are already limiting their use of AI; for AI agents specifically, about one in ten report cost constraints.
The cost also lands on vendors. ServiceNow’s non-GAAP subscription gross margin fell to 80.5% in Q2 2026 from 83% a year earlier, and the company tied its full-year gross margin guidance partly to an acceleration of customer AI adoption (July 22, 2026). Cheaper tokens did not prevent margin compression at one of the largest agent platforms; they made more usage affordable, and usage won.
What this means for a 2027 budget: price agents per completed task, not per token. A token price cut of 10x means little if an agent workflow uses 20x more tokens than the chatbot it replaced.
What Could Change This Outlook?
| Signal | Date | Pushes toward | Why it matters |
|---|---|---|---|
| Gartner’s next quarterly AI spending update | next quarterly release | Upside if 2027 rises again; downside if trimmed | Resets the $29.2B starting point and the $65.5B comparison |
| Salesforce fiscal Q3 2027 results (quarter ended Oct 31, 2026) | next earnings release | Upside if Agentic Work Units keep growing near 97% a quarter | Direct read on usage-based agent wallets |
| End-2026 check on Gartner’s “40% of enterprise apps with task-specific agents” prediction | December 31, 2026 | Downside if well short | First grade-back of a major agent adoption forecast |
| McKinsey State of AI 2027 | expected 2H 2027 (2025 edition: November; 2026 edition: August) | Upside at ≥53% large-org scaling; downside at ≤45% | Directly tests the Axis adopter assumption |
| EU AI Act Annex III high-risk obligations apply | December 2, 2027 | Downside for EU deployments in hiring and credit if compliance costs delay rollouts | Regulatory cost lands inside the forecast year |
| End-2027 check on Gartner’s 40% project-cancellation prediction | December 31, 2027 | Downside if cancellations exceed 40% | Tests the slowdown built into the base case |
Methodology
What was included. Every forecast in the reconciliation table was read at its publisher’s own page or press release on October 28, 2026. Forecasts are copied at the precision published. Where a firm states only a horizon value, no intermediate year is shown.
What was excluded. A widely repeated figure of $201.9 billion in 2026 “agentic AI” spending attributed to Gartner’s January 2026 forecast could not be found in Gartner’s January 15, 2026 press release and is excluded until located in a Gartner document. A claim circulating on vendor blogs that 45% of Fortune 500 companies run production agents, attributed to McKinsey, does not appear in McKinsey’s 2026 survey and is excluded.
How CAGRs were derived. (end value ÷ start value)^(1 ÷ years) − 1, using each firm’s own start and end values.
Base Case limitations. The model combines a survey share (McKinsey, GDP-weighted, self-reported) with a dollar forecast (Gartner). It uses the survey only as a growth ratio, never as a level, but the two methods do not align perfectly: McKinsey fields in mid-year, Gartner measures calendar-year spending. The spend-per-adopter factor is inferred from one year of history. The model excludes consumer agent spending by construction, which Gartner’s segment now includes. Gartner’s September 16, 2026 figures were read from Gartner’s newsroom listing and corroborated by CIO.com’s September 23, 2026 coverage; the full page blocked automated retrieval.
Changes between editions. Any change to drivers or assumptions in later editions will be logged here, and past projections will not be rewritten.
Grade-back commitment. We will compare these projections with reported outcomes in our November 2027 edition, using Gartner’s latest 2027 estimate and McKinsey’s 2027 survey, and again once full-year 2027 figures are published. Next year’s edition opens with that comparison, right or wrong.
About This Dataset
The file ai-agents-trends.csv contains the forecast reconciliation table (nine published forecasts, seven firms), the Axis Base Case inputs and scenario outputs, and every leading-indicator figure cited on this page, with source URL, publication date and retrieval date per row. License: CC BY 4.0. Cite as: Axis Intelligence Research, “AI Agents Trends 2027: Enterprise Adoption, Costs and the Spending Forecast,” 2026.
APA: Axis Intelligence Research. (2026, October 28). AI agents trends 2027: Enterprise adoption, costs and the spending forecast. Axis Intelligence. https://axis-intelligence.com/ai-agents-trends/
AI Agents in 2027: Frequently Asked Questions
How big will the AI agent market be in 2027?
Axis Intelligence Research projects about $44 billion of worldwide spending on AI agents and assistants in 2027 in its base case, with a range of $35–54 billion. Gartner’s September 2026 forecast for the same segment is $65.5 billion, a figure that now includes consumer agents.
What percentage of companies are using AI agents?
It depends on company size. McKinsey’s 2026 survey found 40% of organizations with revenue above $1 billion scaling AI agents, against 22% of smaller ones. “Scaling” means expanding agents beyond a pilot in at least one function; enterprise-wide use is rarer, at about two in ten respondents.
Why do AI agent market size estimates differ so much?
Scope. Standalone studies (MarketsandMarkets, BCC Research, Grand View Research, Fortune Business Insights, Precedence Research) size 2025 at $7.29–8 billion. Gartner’s broader agents-and-assistants segment puts the same year at $16.5 billion.
Are most agentic AI projects failing?
Not most — but a large minority is expected to be dropped. Gartner predicts over 40% of agentic AI projects will be canceled by the end of 2027 because of cost, unclear value or weak risk controls.
Will AI agents replace SaaS?
Partly, at the edges. 32% of McKinsey’s 2026 respondents say their organization skipped at least one software purchase because coding agents let them build it internally. Systems of record are less exposed: Gartner expects incumbent software providers to be the main channel for selling AI to enterprises in 2026.
When do EU AI Act rules apply to AI agents used in hiring or credit?
From December 2, 2027, for stand-alone high-risk (Annex III) systems, after the Digital Omnibus on AI (Regulation (EU) 2026/1744) deferred the original August 2026 date. Agents used in employment or creditworthiness decisions are the most likely to fall in scope.