Home/Blog/AI ROI in 2027
Updated November 2026

AI ROI in 2027

By Arsh Singh/November 2026/10 min read

AI ROI is the measurable business return you get from AI spend, after you count licenses, integration, training and the management time it takes to change how people work. For a 50 to 1,000 person company, that return depends far more on how many people actually change their daily work than on which tool you picked.

For illustration: imagine a 200-person company where the COO approves 120 AI assistant licenses in January. By June, the CFO asks what the company got for the money. Nobody can answer, because nobody set a baseline and nobody owns the number. The tool was fine. The plan around it did not exist. I have strong opinions about this, and I will state the main one now: most disappointing AI results are management problems wearing a technology costume. This post shows you where you probably stand and what tends to go wrong at the next stage. The full sequencing is something I do with clients inside an engagement, and I will say where that line sits.

Key takeaways

  • 88% of organizations use AI in at least one function, yet only about 39% report any enterprise-level EBIT impact (McKinsey, 2025).
  • Redesigning workflows had the biggest effect on bottom-line impact from gen AI among the attributes McKinsey tested, and only 21% of gen AI users had fundamentally redesigned at least some workflows (McKinsey, 2025).
  • Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025 (Gartner, 2024).
  • Adoption is the multiplier. The same license bill can cost 3x more per active user when only a third of staff use the tool.

What is a realistic AI ROI for a mid-size company?

A realistic AI ROI is one you can defend to your CFO with a baseline, an owner and active-use data. The honest market picture is that most companies cannot yet do that, so the typical return is unproven rather than negative.

Start with the spread between usage and results. McKinsey reports that 88% of organizations use AI regularly in at least one business function, while only about 39% report any EBIT impact at the enterprise level, and nearly two-thirds have not begun scaling AI across the company (McKinsey, 2025). MIT's NANDA initiative went further in its 2025 report, finding that 95% of organizations saw no measurable P&L return from enterprise gen AI pilots, as covered by Fortune (2025). Read those two numbers together. Almost everyone is using AI, and few can show money.

Why the gap? The mechanism is simple. A license is a cost the day you buy it. The benefit only arrives when a person does a task differently, and that person's manager lets the new way stick. Skip either step and you have bought a subscription, not a capability.

Here is a worked example, clearly hypothetical. For illustration: 120 licenses at $30 per user per month cost $43,200 a year. If 120 people use the tool weekly, that is $360 per active user. If only 40 do, it is $1,080 per active user, three times the price for the same invoice. Now add time. Say each active user saves 2 hours a week over 48 weeks, at a loaded cost of $45 an hour. That is $4,320 of time per person. Forty users give you $172,800 of capacity; 120 users give you $518,400.

Those hours are not cash yet. Hours saved only becomes ROI when the time is redeployed into revenue work, absorbed growth without new hires, or removed overtime. I have seen this error sink more business cases on paper than any technical failure could. Capacity that nobody redirects just disappears into the working day.

How do you measure AI ROI before you spend more?

You measure AI ROI by fixing a baseline for one workflow, naming one owner, tracking active use rather than seats, and converting saved time into a decision your finance team accepts. Do this before you approve a second tool.

The general approach has six steps. I keep the scoring model and the sequencing for client engagements, but these steps are enough to stop the worst waste.

  1. Pick a workflow, not a tool. "Customer onboarding takes 9 days" is measurable. "Roll out an assistant" is not.
  2. Record the baseline first. Cycle time, error rate, cost per transaction, whatever the workflow already produces. If you cannot measure it today, you cannot claim a gain in six months.
  3. Name a business owner, not an IT owner. The person who runs the workflow owns the result. IT owns access and security.
  4. Track active use weekly. Licenses assigned is a vanity number. People who used the tool on the workflow this week is the number that predicts return.
  5. Decide in advance what saved time becomes. More volume, faster response, fewer contractors, or a role redesign. Write it down.
  6. Set a 90-day review with a stop rule. If active use is under an agreed line, you fix adoption or you stop. No third option.

Step four deserves a real-world parallel from our own work. In the EarningsCall campaign, a US fintech app for investors, 60% of downloads were monthly active, which meant 3K+ MAU on 5K+ downloads, and the product passed 100K+ downloads later (EarningsCall case study). Enjoy Movies Your Way, a US family streaming filters app, reached 38K+ monthly active users on 50K+ downloads by December 2023, a 76% active rate (Enjoy Movies case study). Those are app products, not AI deployments. The lesson transfers anyway: installs and seats flatter you, and active use tells you whether anything real happened. The same discipline applies to a license count.

Where do you stand right now?

Most 50 to 1,000 person companies sit at one of four stages. Each has a typical next failure.

  • Stage 1, Buying: licenses are approved, no baseline exists. Next failure: the six-month "what did we get?" meeting with no answer.
  • Stage 2, Piloting: a few enthusiasts use tools, results are anecdotes. Next failure: the pilot succeeds in one team and dies when it meets a manager who was never briefed.
  • Stage 3, Scaling: several teams use AI, no shared rules. Next failure: a data or compliance incident forces a freeze.
  • Stage 4, Embedded: workflows are redesigned, owners report numbers. Next failure: costs creep as usage grows and nobody reviews the portfolio.

If you read that list and recognize two stages at once, good. That is normal. Different departments move at different speeds, and a leader's first job is to see the map.

The data says adoption and governance decide AI ROI

Across the largest published surveys, the pattern is consistent: companies that redesign workflows and put senior leaders in charge of governance report better financial results, and companies that stop at tool purchase report little. AI governance is the set of owners, rules and review points that decide which AI use is allowed, who is accountable for it, and how its risks and costs are checked.

McKinsey's March 2025 survey found that redesigning workflows had the biggest effect on EBIT impact from gen AI among the attributes it tested, and that CEO oversight of AI governance was the element most correlated with higher self-reported bottom-line impact (McKinsey, 2025). Only 21% of gen AI users had fundamentally redesigned at least some workflows. So the majority are layering a new tool on an old process and hoping.

Gartner's warnings point the same way. It predicted at least 30% of gen AI projects would be dropped after proof of concept, citing poor data quality, inadequate risk controls, escalating costs or unclear business value (Gartner, 2024). A year later it predicted over 40% of agentic AI projects would be canceled by the end of 2027 for similar reasons (Gartner, 2025). Notice what is missing from those causes: "the model was not good enough."

SourceFindingNumberWhat it means for a 50 to 1,000 person firm
McKinsey, 2025Organizations reporting enterprise EBIT impact from AI39%Use is common; proven return is not.
McKinsey, 2025Gen AI users that fundamentally redesigned some workflows21%Workflow redesign is the lever most firms skip.
Gartner, 2024Gen AI projects predicted to be abandoned after proof of concept30%+Pilots need an exit or scale decision, not drift.
Gartner, 2025Agentic AI projects predicted to be canceled by end of 202740%+Agents raise the cost and risk bar; governance first.
Fortune on MIT NANDA, 2025Organizations with no measurable P&L return from gen AI pilots95%Without a measured baseline, "no return" is the default.

The synthesis I draw from this table: three independent sources land on the same cause, which is that value comes from changing the work and managing the risk, not from model access. If you are a mid-size company, you have an advantage here. A 300-person firm can redesign a workflow in weeks. A 30,000-person firm needs a program office. If you want help turning this into a plan for your teams, our AI automation and implementation service starts with exactly this workflow and ownership review.

Why do AI projects stall after the pilot?

AI projects stall after the pilot because the people who must change their daily work, mostly middle managers, were never trained to lead that change. The tool works in a demo; the team keeps using the old process because the manager still rewards it.

This is the pattern I would bet on in almost any mid-size company. Tools get bought at the executive level. Staff get a one-hour webinar. Managers get nothing, so they cannot answer the questions that decide adoption: Which tasks should my team do with this? What do I stop asking for? How do I judge the quality of AI-assisted work? When those answers are missing, the safest move for a manager is to carry on as before. That is rational behavior, and it kills ROI quietly.

Here are the mistakes I would check for first, each with a short illustration.

  1. Counting licenses as adoption. Imagine a company reporting "100% rollout" because every employee has a seat. Weekly active use is 25%. The dashboard is green and the business case is red.
  2. Claiming hours saved as dollars saved. Imagine a finance team that saves 15 hours a week and spends them on more meetings. The cost line has not moved.
  3. Training staff but not managers. Imagine a customer support team trained on an assistant, led by a supervisor who still scores agents on handle time alone. Agents drop the tool within a month because the metric punishes the extra review step.
  4. Letting each department buy its own tools. Three teams pay for three overlapping products, and nobody can say which data went where. That is how a scaling company ends up in a compliance freeze.
  5. Treating governance as a legal document. A policy PDF nobody reads is not governance. Governance means a named owner, an approved-use list, and a review date.
  6. Skipping HR. Role descriptions, performance criteria and training budgets are HR's levers. If HR is not in the room, the workflow changes on paper only.

If your managers cannot explain what changes in their team's week, you have not deployed AI. You have distributed software.

One more point for operations leaders. When a pilot succeeds, resist the urge to scale it to everyone immediately. Scale to the teams whose managers have been briefed and whose baseline you measured. A smaller rollout with a clean before-and-after beats a company-wide rollout you can never evaluate.

What will AI ROI look like in 2027?

In 2027, AI ROI will be judged like any other capital decision: with baselines, owners and kill criteria. The era of funding AI on enthusiasm is closing, and finance teams are starting to ask for proof.

Three shifts look likely based on where the published data points. First, scrutiny rises. If Gartner's prediction holds and over 40% of agentic projects are canceled by the end of 2027 for cost, unclear value or weak risk controls, boards will ask harder questions about every agent proposal. Companies with measured baselines will get approvals; those without will wait.

Second, training moves from optional to expected. The EU AI Act's Article 4 asks providers and deployers of AI systems to take measures to ensure a sufficient level of AI literacy among their staff. If you employ people in the EU or sell there, HR will own part of this. Even outside the EU, I expect customers and insurers to ask what training you give.

Third, the manager becomes the unit of change. The firms that pull ahead in 2027 will not be the ones with the most tools. They will be the ones where a team lead can say what AI does in their team's week, how quality is checked, and what the number was before. Build that skill now, and a tool swap next year costs you little. Skip it, and every new tool repeats the same stall.

Frequently Asked Questions

How long does it take to see AI ROI?

For a single, well-defined workflow with a baseline, you can usually judge direction within 90 days. Company-wide returns take longer because they depend on workflow redesign and manager adoption. I advise clients to treat the first 90 days as a test of adoption, not of savings, and to set a stop rule in advance.

What is the biggest reason AI ROI disappoints?

Lack of adoption tied to unchanged workflows. The sources above point to weak workflow redesign, unclear business value and poor risk controls, not weak models. My view is that the tool is rarely the problem; the missing piece is a manager who knows what to change and a baseline to prove it did.

Should HR be involved in AI implementation?

Yes, from the start. HR controls training budgets, role definitions and performance criteria, which decide whether people actually change how they work. If HR joins only at rollout, you get a policy and a webinar. If HR joins at design, you get managers who are trained and measured on the new way of working.

How do I calculate ROI on an AI assistant license?

Divide total annual cost (licenses, training, admin time) by weekly active users, then compare it with the value of redeployed time or avoided spend. Do not count hours saved unless you can say what the time became. As a rough guide, if active use is a third of seats, your effective cost per user triples.

Is AI governance only for large companies?

No. A 100-person company needs it too, in lighter form: one accountable owner, a list of approved tools and data types, and a review date. McKinsey's survey tied senior oversight of AI governance to better reported bottom-line impact, and smaller firms can set that up in weeks rather than quarters.

Conclusion

AI ROI comes down to four principles. Measure a workflow before you buy a tool. Track active use, because seats flatter you. Turn saved time into a decision, because hours alone are not money. Train managers to lead the change, because they decide whether the work really changes.

If you read the four stages above and recognized yourself, you already know your next likely failure. What to fix first, and in what order, depends on your teams, your data and your risk exposure, and that is what I work through with leaders in an engagement. If you want a straight read on where your company stands, book a free strategy call and bring your current tools, your headcount and your toughest workflow.

Want a second pair of eyes on your growth?

Book a free 30-minute strategy call. Bring your numbers, leave with two or three moves worth making. No pitch, no deck.

Book a Free Strategy Call