Source: https://frntir.ai/guides/agentic-ai-roi-by-industry

Sourced agentic AI ROI by industry (finance, healthcare, legal, manufacturing, retail, HR), with realistic time-to-value and which numbers to trust.

---

Data & Guide

# Agentic AI ROI by Industry: What Mid-Sized Businesses Can Realistically Expect

By **Aidan Dunphy** & **Cyril Le Roux**, Co-founders of Frntir · Last updated: **October 2026** · 16 min read

AI returns are real, but they are wildly uneven. Averaged across the market the return is roughly $3.70 for every $1 invested, yet independent research shows only a minority of firms can point to hard bottom-line impact. This guide sets out what ROI looks like sector by sector, how long it takes, and which numbers you can trust.

The short version

- **The averages are good, the distribution is skewed.** About $3.70 back per $1 on gen AI, but the value is concentrated in the firms that did the groundwork.
- **The biggest wins share a shape.** High-volume, document-heavy, rule-based work: underwriting, contract review, clinical notes, demand forecasting.
- **Time-to-value ranges from weeks to a couple of years.** Narrow tasks pay back in weeks to months; organisation-wide value takes one to three years.
- **Trust the sourced numbers, not the calculators.** Many headline sector stats trace only to vendor marketing. We have left those out.

## What ROI can you realistically expect from AI?

Across the market, IDC’s 2024 study (sponsored by Microsoft) found an average return of about $3.70 for every $1 invested in generative AI, with top performers nearer $10.30, and value realised within roughly 13 months (IDC, 2024). That is the optimistic, vendor-sponsored headline. The counterweight: independent research finds strong returns are still the exception, not the rule.

$3.70

average return for every $1 invested in generative AI, with top performers nearer $10.30

IDC / Microsoft, 2024

<8 months

typical time to deploy an AI project, with value usually realised within about 13 months

IDC / Microsoft, 2024

~39%

of organisations attribute any EBIT impact to AI so far, and most of those put it under 5%

McKinsey, 2025

~15%

of enterprises report significant, measurable ROI from gen AI, though 74% say their top initiative is meeting expectations

Deloitte, 2024

Put the promise and the reality side by side and the picture is clear. McKinsey’s 2025 State of AI found that only about 39% of organisations attribute any EBIT impact to AI so far, and most of those put it at under 5% of EBIT, with roughly 6% counting as high performers (McKinsey, 2025). Deloitte’s 2024 State of Generative AI found that while about 74% say their most advanced initiative is meeting or exceeding ROI expectations, only about 15% report significant, measurable ROI (Deloitte, 2024). Both things are true at once: the ceiling is high, and most firms have not reached it yet.

## Why ROI varies so much by sector

The sector matters less than the shape of the job. Returns cluster wherever the work is high-volume and rule-based. A repetitive, document-driven task pays back faster than a bespoke, judgement-heavy one, whatever industry it sits in.

This is why IDC (2024) ranks financial services first on gen-AI returns: underwriting, claims and onboarding are structured, repetitive and expensive, so a small percentage improvement is worth a lot. The same logic explains why legal document review and clinical documentation show up strongly, and why sectors dominated by one-off, relationship-heavy work show thinner, slower returns. When you read the table below, look less at the industry label and more at whether the winning use case is a job your own team repeats often enough to matter.

## Agentic AI ROI by industry

Here is the sector-by-sector picture, with a sourced result and a realistic time-to-value for each. Every figure is attributed to a named source, and where a source is a vendor-sponsored survey we say so. Figures are indicative: they come from different studies, geographies and years, so use them to set expectations, not to forecast your own numbers.

| Sector | Where the strongest ROI shows up | A sourced result | Typical time to value |
| --- | --- | --- | --- |
| Financial services & insurance | Underwriting, claims processing, KYC/AML onboarding, fraud detection | Hiscox cut lead underwriting on one specialist book from about 3 days to about 3 minutes (Hiscox, 2024). Allianz reports roughly an 80% cut in claim processing and settlement time on its agentic claims workflow (Allianz, 2025). | Value ~13 months; the highest gen-AI ROI of any sector (IDC, 2024) |
| Healthcare | Clinical documentation (ambient scribes), prior authorisation, coding, scheduling | The Permanente Medical Group saved roughly 15,700 documentation hours in a year across 7,260 physicians (Kaiser Permanente / NEJM Catalyst, 2024 to 2025). An independent multi-site study found a more modest saving of about 16 minutes per 8-hour shift (STAT, 2026). | Value ~14 months; documentation wins land fastest |
| Legal & professional services | Contract and NDA review, due diligence, risk identification | In a Luminance client case, Bird & Bird reviewed about 200,000 documents in two weeks with two associates, lifting throughput from 79 to 3,600 documents an hour (Luminance / Global Legal Post, 2018). | 2 to 8 weeks on a well-scoped review workflow |
| Manufacturing & logistics | Predictive maintenance, supply-chain and inventory optimisation | Predictive maintenance typically raises equipment uptime 10 to 20% and cuts maintenance costs 5 to 10% (Deloitte, 2017). AI supply-chain early adopters cut logistics costs 15% and inventory 35% (McKinsey, 2021). | 3 to 6 months to validated ROI on a maintenance or supply-chain pilot |
| Retail & e-commerce | Demand forecasting, inventory, customer service | AI demand forecasting reduces forecasting errors 20 to 50% and lost sales by up to 65% (McKinsey, 2021). On gen-AI returns, retail sits mid-pack, behind financial services (IDC, 2024). | 6 to 18 months for forecasting models to prove out |
| Recruitment & HR | CV screening, sourcing, interview scheduling, onboarding | In Workable’s survey of AI-using hiring teams, about 85% reported meaningful time savings (Workable, 2024). Hard per-hire savings, though, are mostly vendor-calculator estimates, so treat them with caution. | 1 to 3 months on narrow screening tasks |

A few of these deserve a note. In financial services, the widely-repeated "20x speed, 80% cost" claims for claims processing trace only to vendor marketing, so we have used named cases instead: Hiscox’s underwriting result and Allianz’s reported cut in claim processing time (Hiscox, 2024; Allianz, 2025). The 77% of financial-services firms said to report positive returns within a year comes from a Google Cloud survey, so read it as a vendor-sponsored figure (Google Cloud / National Research Group, 2024).

Healthcare is the clearest example of why the counterweight matters. The Permanente Medical Group’s roughly 15,700 saved documentation hours is a large, independently-reported result (Kaiser Permanente / NEJM Catalyst, 2024 to 2025), but an independent multi-site study found the per-clinician saving was far smaller, about 16 minutes per 8-hour shift (STAT, 2026). Both are real. The lesson is that aggregate hours saved across thousands of users can be large even when the per-person minute-level saving is modest.

In legal and professional services, the Bird & Bird case is genuine but comes from the vendor whose tool was used, so treat "79 to 3,600 documents an hour" as a vendor-reported client result (Luminance / Global Legal Post, 2018). The scale of investment is easier to stand up: Deloitte has committed around $3bn, KPMG around $2bn and EY around $1.4bn to AI (Bloomberg Tax, 2024 to 2025). In manufacturing and logistics, we have deliberately dropped the often-quoted "General Motors saves $20m a year" line, which we could not trace to any GM source, in favour of Deloitte’s and McKinsey’s sourced ranges, plus BCG’s finding that supply-chain AI can add roughly 2 to 4 percentage points of EBITDA (BCG, 2024). In HR, per-hire pound-savings figures almost always come from vendor ROI calculators, so we have kept only the attributed survey number.

## How to read AI ROI numbers without getting burned

Most of the damage in AI business cases comes from taking a vendor’s best-case number as a planning assumption. A simple filter: if a figure has no named source, or the only source is the company selling the tool, treat it as a ceiling to aim at, not a baseline to budget on.

- **Separate independent research from vendor surveys.** McKinsey, Deloitte and government studies are measuring the whole market. A vendor’s case study is measuring its best customer. Both are useful, but only one tells you the odds.
- **Watch the currency and the year.** The much-quoted "$3.50 per $1" is IDC’s older figure; the current one is $3.70 (2024). A "£3.20" healthcare return is US$3.20 misquoted. Small errors like these are how a stat drifts away from its source.
- **Prefer aggregate, real-world deployments.** "Saved 15,700 hours across 7,260 physicians" is more trustworthy than "saves 3 hours a day": it is measured at scale rather than demoed once.
- **Assume the average firm, not the top performer.** The gap between the $3.70 average and the $10.30 top-performer figure is the groundwork. If you have not done it, plan for the average.

## What determines your ROI

The sector benchmarks set the ceiling. Whether you get near it is decided by the same four factors every time: a job worth automating, clean-enough data, a fixed process, and people who adopt the tool. The technology is rarely the deciding variable.

This is the single most consistent finding in the research, and it is why we wrote a whole guide on [why most AI projects fail](https://frntir.ai/guides/why-ai-projects-fail). The short version: roughly 70% of what makes AI succeed is people and process, not the algorithm (BCG, 2024). The firms that hit the top-performer returns pick one contained, high-value job, map how it works, fix the broken parts before automating, measure a baseline, and scale only what works. The firms that see nothing usually bought a tool before they had defined the job. If you want the full playbook, start with [that guide](https://frntir.ai/guides/why-ai-projects-fail), then come back to the sector table to size the opportunity.

## How AI colleagues change the ROI math

If most of the return depends on adoption and on knowledge accumulating rather than leaking away, the shape of the tool matters. That is the idea behind Frntir’s [AI Synths](https://frntir.ai/guides/ai-agents-vs-copilots-vs-chatbots): a named AI colleague with persistent memory that owns a recurring job, rather than another system your team has to remember to open.

A Synth works inside the tools your team already uses, operates with bounded, explainable autonomy under a human boss, and keeps a full audit trail. That targets the two things the numbers above keep coming back to. It lowers the adoption barrier by living where people already work rather than being a new destination, and it compounds as the context it builds up (past decisions, what worked, who to ask) stays with the business instead of walking out with a spreadsheet. It does not remove the need to pick a good job and fix the process first; nothing does. But on the jobs where AI already pays back, a colleague that owns the whole loop tends to close more of the gap between the average return and the top-performer return.

See it in practice

Vision Meditech, a 25-year manufacturer, hired a Synth named Wallace and started with one contained, high-value workflow: complex quote replies. Those went from an hour to minutes, and the hardest feasibility answers from three weeks to the same day. That is the pattern the sector data points to, narrow, measurable, repeated. [Read the Vision Meditech case study](https://frntir.ai/case-studies/vision-meditech).

## Frequently asked questions

What is the average ROI on AI? The most-cited benchmark is IDC’s 2024 study (sponsored by Microsoft), which found an average return of about $3.70 for every $1 invested in generative AI, with top performers nearer $10.30, and value typically realised within about 13 months. Treat that as an optimistic, vendor-sponsored figure. Independent surveys are soberer: McKinsey (2025) found only about 39% of organisations attribute any EBIT impact to AI so far, and Deloitte (2024) found only about 15% report significant, measurable ROI. In short, strong returns are real but concentrated in a minority of firms that did the groundwork.  How do you calculate ROI on an AI project? ROI is the net value the project returns divided by what it cost, written as (benefit minus cost) over cost. For AI the benefit is the hours, error rate or cycle time it removes, converted to money; the cost is licensing, build, integration and the change management to get people using it. The formula is the easy part. The work is setting a solid baseline for both sides before you start, so you measure a real change rather than a hoped-for one.  What costs should we include in an AI ROI calculation? Include every cost, not just the licence. A realistic total covers the software or usage fee, the implementation and integration work, data clean-up, the staff time to configure and adopt it, and the ongoing cost to run and maintain it. Usage-priced AI adds a variable line that grows with volume, so model it at your expected load rather than a demo’s. Leaving out change management is the most common reason a business case that looked positive on paper comes in flat.  What is a realistic first-year ROI for a mid-sized business? Modest, for most firms. The headline averages (about $3.70 back per $1, IDC 2024) describe the firms that did the groundwork, while independent research finds only about 15% report significant, measurable ROI and about 39% can attribute any EBIT impact at all (Deloitte 2024; McKinsey 2025). A fair year-one plan is to prove payback on one contained job and treat business-wide returns as a one-to-three-year effort, not a first-year line.  What is a good target ROI to aim for on an AI project? Aim for a clear, measured payback on one job rather than a headline multiple. The quoted averages (about $3.70 per $1, top performers nearer $10.30, IDC 2024) are a useful ceiling, but a mid-sized business is better served by a concrete target: cut this job’s cycle time or cost by a set percentage within the pilot window, measured against a baseline. A real, modest, proven number is worth more than an impressive projected one.  How long does AI take to pay back? IDC (2024) reports AI projects deploying in under 8 months and realising value within about 13 months on average. Narrow, well-scoped tasks are faster: legal document review can show value in 2 to 8 weeks, HR CV screening in 1 to 3 months. Organisation-wide value is slower, typically one to three years of sustained effort. Be wary of any vendor promising quarterly returns across the whole business.  How long should an AI pilot run before we judge its ROI? Long enough to clear the learning curve and build a baseline, which usually means weeks for a narrow task and a quarter or more for anything touching several teams. Judge too early and you measure the setup friction; judge too late and you have spent on something that was never going to work. Set the baseline and the success threshold before the pilot starts, then run it until adoption is steady rather than stopping on the first good or bad week.  What KPIs should we track to prove AI ROI? Track the ones you can baseline before the project starts: cycle time per task, cost per task, error or rework rate, throughput, and adoption (how many people use it, how often), plus revenue or margin where the job touches them. The most important step is recording those numbers before go-live. An AI ROI claim with no pre-project baseline is a guess dressed as a measurement, and finance will treat it as one.  How do we measure AI ROI when the saving is time, not revenue? Convert the time saved into its cost, but measure it where the saving actually lands. Aggregate hours across a team can be large even when the per-person saving is small: ambient scribes saved about 15,700 hours across 7,260 physicians, yet an independent study put the per-clinician gain at roughly 16 minutes a shift (Kaiser Permanente / NEJM Catalyst 2024 to 2025; STAT 2026). Price the aggregate, and count only hours that turn into lower cost or more output.  How do we build an AI business case that finance will accept? Baseline first, then size the opportunity against named sources rather than vendor calculators. Measure how the job performs today (time, cost, error rate), use independent benchmarks as a ceiling to aim at rather than a figure to budget on, and plan for the market average unless you have done the groundwork top performers did. A controlled pilot with a before-and-after on one job gives finance a figure to extrapolate from, which beats a projection built on a supplier’s best case.  Which industries get the most ROI from AI? On generative-AI returns, IDC (2024) puts financial services first, with retail and consumer goods mid-pack. In practice the biggest wins cluster wherever work is high-volume, document-heavy and rule-based: underwriting and claims in insurance, contract and due-diligence review in legal, clinical documentation in healthcare, and demand forecasting in manufacturing and retail. The sector matters less than the shape of the job. A repetitive, measurable, document-driven task pays back faster than a bespoke, judgement-heavy one, whatever industry it sits in.  Which business functions tend to see AI ROI first in a mid-sized business? The back-office, document-driven functions. Finance and procurement (invoice and quote handling, spend checks), customer service (triage and first-response drafting), HR (screening and scheduling), and any bid, proposal or knowledge work where the same document tasks repeat. These are where a mid-sized team has volume without enterprise complexity, so a contained tool can prove a return quickly before anything wider is attempted.  Which AI use cases have the fastest payback? High-volume, document-heavy, rule-based jobs. The work that pays back quickest is repetitive and measurable: contract and due-diligence review, clinical documentation, claims and underwriting, CV screening, demand forecasting. These share a shape, a task the team already does many times a day where a small percentage gain compounds. Bespoke, judgement-heavy or relationship-led work pays back slowly if at all, whatever sector it sits in.  Is AI ROI only worth it at enterprise scale, or can a small team see returns? A small team can see strong returns on a single high-value job. The headline studies skew towards large enterprises (that is who gets surveyed), yet the returns come from the shape of the task, not the size of the firm. One recurring, measurable, document-heavy workflow can pay back for a 30-person team as clearly as a 3,000-person one. What does not work at small scale is buying a broad platform before you have defined the job it is for.  Does AI ROI grow over time or plateau? It can do either, depending on whether knowledge accumulates. Returns compound when the tool keeps context (past decisions, what worked, who to ask) so each month starts ahead of the last. They plateau, or fade, when adoption stalls or the knowledge leaves with the person who held it. A tool that owns a recurring job and retains its context tends to keep improving; a tool people forget to open does not.  What is the ROI of agentic AI compared with a chatbot or a copilot? The returns differ with how much of the job the tool does. A chatbot answers, a copilot assists a person task by task, and an agentic system owns a recurring job end to end under human oversight. More of the work moved off the team usually means more measurable return, and also more that has to be set up correctly and trusted. Our guide on <a href="/guides/ai-agents-vs-copilots-vs-chatbots">AI agents, copilots and chatbots</a> draws the distinction before you size the return.  How is AI ROI different from traditional automation or RPA? Traditional automation (RPA) follows fixed rules and pays back on stable, structured, high-volume tasks, but it breaks when the process changes. AI, and agentic AI in particular, handles variation and language, so it reaches jobs RPA could not (reading a messy quote, drafting a tender response) and copes better when inputs shift. The trade-off is that AI output needs checking and a clear owner, so the ROI case rests on how well you supervise it, not on volume alone.  How much does it cost to run an AI agent or AI colleague? It varies by pricing model, and the model matters as much as the headline rate. Per-seat pricing is predictable; consumption or outcome pricing scales with how much work the system does, which is harder for a mid-sized team to forecast and can surprise you at volume. Model any usage-based quote at your real expected load. Frntir prices Synths on a custom basis against the job they own rather than per seat, so the cost tracks the work rather than headcount.  How does the ROI of AI compare with hiring someone for the same job? It is a fair comparison to run, and the answer is job-specific. An AI colleague is always on, carries no recruitment or attrition cost, and keeps an audit trail, which suits a recurring, high-volume job. A person is the right choice where the work is judgement-heavy, relationship-led or changes constantly. The strongest case often blends the two: automate the repetitive core of a role so the person spends their time on the part that needs a person.  What data do we need in place before AI delivers ROI? Clean-enough data for the specific job, not perfect data everywhere. What matters is whether the information the task depends on is accessible, reasonably accurate and not scattered across systems nobody reconciles. You do not need a data-warehouse programme before you start; you need the inputs for one job to be good enough that the output can be trusted. Fixing the data for a single workflow is usually faster and cheaper than the automation it unlocks.  Are these AI ROI statistics reliable? Read them carefully. A large share of the eye-catching sector numbers (per-hire savings, "10x" customer-service returns, "99% faster" claims) trace only to vendor marketing or ROI calculators, not to independent studies, so we have left those out of this guide. The figures here are attributed to named sources and flagged where the source is a vendor-sponsored survey. The most reliable single data point is the reality check: independent research (McKinsey 2025, Deloitte 2024) shows most organisations have not yet proven hard bottom-line ROI, even as headline case studies look spectacular.  Can we trust vendor ROI calculators and case studies? Use them for direction, not for budgeting. A vendor case study measures its best customer and a calculator models its best case, so both show the ceiling rather than the odds. Independent research (McKinsey, Deloitte, government and academic studies) measures the whole market, including the firms that saw little. Weigh a supplier’s showcase against that market picture, and if a figure has no named source beyond the company selling the tool, treat it as marketing.  How do we avoid paying for AI ROI that never materialises? Treat every unsourced or vendor-calculated figure as a ceiling, not a plan, and never buy a platform before you have defined the job. The firms that see nothing usually bought a tool first and went looking for a use for it. Run a small, measured pilot on one job you have baselined, and scale only what beats the baseline. If a supplier cannot point to an independent result on a job like yours, assume the market-average return, not their showcase number.  How do we get ROI from AI in a mid-sized business? Pick one recurring, high-value job whose outcome you can measure, map how it works, fix the broken parts before automating, set a baseline, and run a small controlled pilot before scaling. The sector case studies that deliver ROI almost always start narrow. Most failures come from people, process and data problems rather than the technology, which is covered in our guide on why most AI projects fail. Choose a job first, then a tool, not the other way round.  Who should not expect strong AI ROI yet? Firms whose work is bespoke, judgement-heavy and rarely repeated, those without clean-enough data for the target job, and those with no capacity to change how people work. If you cannot name one recurring job, measure its baseline and get people to adopt a new way of doing it, the technology will not rescue the business case. For some teams the better move is to fix the process and the data first, then revisit AI once there is a defined job for it to own.

## Sources

1. IDC InfoBrief, sponsored by Microsoft, [The Business Opportunity of AI](https://blogs.microsoft.com/blog/2024/11/12/idcs-2024-ai-opportunity-study-top-five-ai-trends-to-watch/) (2024): ~$3.70 return per $1, top performers ~$10.30, deploy in under 8 months, value in ~13 months.
2. McKinsey & Company, [The State of AI](https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai) (2025): ~39% attribute any EBIT impact; ~6% high performers.
3. Deloitte, [State of Generative AI in the Enterprise](https://www.deloitte.com/uk/en/issues/generative-ai/state-of-generative-ai-in-enterprise.html) (2024): ~74% meeting ROI expectations vs ~15% reporting significant measurable ROI.
4. Hiscox, [AI underwriting model](https://www.hiscoxgroup.com/news/press-releases/2024/12-08-24) (2024): lead underwriting on a specialist book from ~3 days to ~3 minutes.
5. Allianz, [Agentic AI in claims (Project Nemo)](https://www.allianz.com/en/mediacenter/news/articles/251103-when-the-storm-clears-so-should-the-claim-queue.html) (2025): ~80% reduction in claim processing and settlement time.
6. Google Cloud / National Research Group, *ROI of AI in Financial Services* (2024): 77% of FS firms reporting positive ROI within a year (vendor-sponsored survey).
7. Kaiser Permanente / The Permanente Medical Group, reported in *NEJM Catalyst* and by the [AMA](https://www.ama-assn.org/practice-management/digital-health/ai-scribes-save-15000-hours-and-restore-human-side-medicine) (2024 to 2025): ambient AI scribes saved ~15,700 documentation hours across 7,260 physicians.
8. STAT News, [independent multi-site study on ambient scribes](https://www.statnews.com/2026/04/01/ai-ambient-scribes-modest-time-savings-clinical-documentation/) (2026): ~16 minutes saved per 8-hour shift.
9. Luminance client case study (Bird & Bird), reported by the [Global Legal Post](https://www.globallegalpost.com/news/bird-amp-bird-deploys-artificial-intelligence-technology-40035917) (2018): ~200,000 documents in two weeks, 79 to 3,600 documents an hour (vendor-reported client result).
10. Bloomberg Tax, [Big Four AI rollouts](https://news.bloombergtax.com/financial-accounting/big-four-firms-roll-out-ai-that-can-handle-routine-tasks-solo) (2024 to 2025): Deloitte ~$3bn, KPMG ~$2bn, EY ~$1.4bn committed to AI.
11. Deloitte Insights, [Predictive technologies for asset maintenance](https://www.deloitte.com/us/en/insights/industry/manufacturing-industrial-products/industry-4-0/using-predictive-technologies-for-asset-maintenance.html) (2017): uptime +10 to 20%, maintenance cost -5 to 10%.
12. McKinsey & Company, [Succeeding in the AI supply-chain revolution](https://www.mckinsey.com/industries/metals-and-mining/our-insights/succeeding-in-the-ai-supply-chain-revolution) (2021): logistics costs -15%, inventory -35%, forecasting error -20 to 50%, lost sales down up to 65%.
13. BCG, [Unlocking Impact from AI in Supply Chains](https://www.bcg.com/assets/2024/executive-perspectives-unlocking-impact-from-ai-supply-chains-7oct.pdf) (2024): ~2 to 4 percentage-point EBITDA improvement.
14. Workable, *AI in Hiring* survey (2024): ~85% of AI-using hiring teams reporting meaningful time savings (vendor survey).
15. BCG, [Where’s the Value in AI?](https://www.bcg.com/publications/2024/wheres-value-in-ai) (2024): the 10-20-70 rule (roughly 70% of AI success is people and process).

![Aidan Dunphy](https://frntir.ai/images/Aidan-Dunphy-Profile.webp) ![Cyril Le Roux](https://frntir.ai/images/Cyril-Le-Roux-headshot.webp)

**Aidan Dunphy & Cyril Le Roux** are the co-founders of Frntir.

Aidan has 25+ years in product strategy and technology leadership (B.Sc. Mathematics, Executive MBA). Cyril has 20+ years scaling product organisations, including as VP Product at TransferGo (MBA, The Open University). Frntir builds AI Synths for mid-sized businesses.

## Want to know what ROI a Synth could deliver for you?

Tell us the recurring job you’d start with, and we’ll show you how a Synth could own it, inside the tools your team already uses, with a realistic view of the return and the timeline.

[Book a call](https://frntir.ai/contact)