What are AI Predictions for the Next 5 Years: 9 Changes to Expect by 2031

What are AI Predictions for the Next 5 Years: 9 Changes to Expect by 2031
Share

By 2031, an AI assistant could compare hotels, check their cancellation policies, and prepare a booking for you to approve. At work, the same shift could put AI in charge of routine refunds, invoice checks, and support requests. As AI takes on more of those tasks, companies will have to decide where it can act on its own and where a person needs to step in.

Our AI predictions for the next five years cover what that shift could mean for workers, businesses, and the public. Some changes are already visible in search results and workplace tools. Others, such as robots that can handle a wider range of warehouse tasks, depend on whether the technology proves reliable outside a pilot.

What earlier AI predictions got right

Forrester’s 2017 forecast argued that AI and big data would help businesses use information in more decisions. By 2018, analysts were also paying attention to the work required to make that happen: preparing data, setting rules, and fitting AI into everyday routines. The 2018 AI roundup anticipated more workplace chatbots and a mix of automated tasks and human oversight. Those ideas still apply when a company connects an AI assistant to its own records. Forrester’s 2017 forecast · 2018 AI predictions · 2018 AI, big data, and analytics predictions.

The 2020 AI roundup asked which tasks machines would take over and where people would still make the decisions. The two 2020 cybersecurity roundups flagged privacy, ransomware, and deepfakes. In 2022, Forrester forecast that AI governance would become part of how companies use the technology. Each of those ideas has become more concrete now that AI tools can handle business records, generate convincing audio, and take actions users once performed themselves. 99 AI predictions for 2020 · 141 cybersecurity predictions for 2020 · 42 additional cybersecurity predictions · Forrester’s 2022 outlook for 2023.

1. AI agents will finish more routine work

By 2031, AI agents will handle more tasks with clear rules and a visible endpoint. A customer service agent could check whether an order qualifies for a refund and close a straightforward claim within a set dollar limit. A finance agent could match invoices to purchase orders and send any discrepancies to an employee for review.

There is room for adoption to grow. Stanford’s 2026 AI Index reported that 88% of surveyed organizations used AI in 2025. Deployment of AI agents was still in the single digits across nearly all business functions. The companies that expand their use of agents will need ways to review decisions, catch errors, and reverse actions. Stanford’s 2026 AI Index.

What to watch — The share of routine cases agents close without human intervention and how often employees have to reopen them.

ALSO READ: AI Agents Statistics: Usage And Market Insights (2025 to 2030)

2. AI will change what employers expect from entry-level workers

Junior analysts, developers, and support staff often learn by writing first drafts, fixing simple bugs, or answering common questions. As AI takes on more of that work, some employers may hire fewer people for roles built around those tasks. Others may hire juniors to review AI output, check sources, or investigate unusual cases earlier in their careers.

The International Labour Organization estimated in 2025 that about one in four workers globally held jobs with some exposure to generative AI. The ILO expects changes to existing roles to be more common than entire jobs disappearing because many tasks still require people. By 2031, employers will need to work out how new hires build judgment when software does more of the practice work. ILO research on generative AI and jobs.

What to watch — Changes in the share of job postings open to beginners and in the skills those postings ask for.

3. Reliable company data will make workplace AI more useful

A sales manager can ask an AI assistant why revenue fell last quarter. To answer, the assistant needs accurate records and a clear definition of revenue. If finance and sales count renewals differently, both teams could get plausible answers that lead to different decisions.

By 2031, more companies will build AI into the tools employees already use for business reporting and planning. The most useful systems will let workers see which records the AI used, when those records were last updated, and how it reached a figure. That ability becomes more important when an AI agent can act on the answer.

What to watch — How often workplace assistants show their sources and calculations for answers drawn from company data.

4. Deepfake fraud will change how companies approve payments

4. Deepfake fraud will change how companies approve payments

A voice message that sounds like your boss can make an unusual payment request feel urgent and legitimate. In 2025, the FBI documented a campaign in which attackers used AI-generated voice messages to impersonate senior U.S. officials. The case showed how a convincing voice can help someone gain a target’s trust. FBI alert on voice impersonation.

By 2031, more companies will ask staff to confirm sensitive requests through a separate, trusted channel. An employee who receives a request to change a supplier’s bank details might verify it using a phone number already on file. Security teams could also use AI to group related alerts and summarize incidents, freeing analysts to investigate the cases that need their attention.

What to watch — How many businesses require independent confirmation for bank-detail changes, unusual payments, and account recovery.

5. AI regulation will change how the same app works across countries

Users will see more of AI regulation in the products they open every day. The EU AI Act’s transparency rules already cover disclosures for certain AI interactions and labeling for deepfakes. Which rules apply can depend on what a system does and how people use it. European Commission guidance on AI transparency.

By 2031, someone using an app in the EU could see different notices or features than someone using the same app elsewhere. Companies buying AI tools will also want clearer records of what a system did and who approved the action, especially when it can change records, contact customers, or move money.

What to watch — Product release notes that list country-specific features, disclosures, or approval controls.

6. Smaller models will handle more everyday AI tasks

Sorting support tickets, pulling dates from invoices, and drafting a short reply are narrow jobs. A smaller AI model tested on those tasks may handle them quickly and at lower cost. A request that calls for interpreting several long documents may still need a model with greater capacity.

By 2031, more apps will pick a model automatically based on the work involved. Some simple tasks could run on a phone or laptop, where the hardware allows it. Others will go to the cloud. Users may notice the change in how quickly a feature responds, what it can do offline, or how often they hit a usage limit.

What to watch — How many features run on the device or use a smaller model by default, and how often they need to fall back to a larger one.

7. Electricity will shape where AI capacity grows

AI companies need more than chips and buildings to expand. They also need a reliable supply of electricity. The International Energy Agency has identified slow grid connections as a constraint for some U.S. data center developers. A site can have space for thousands of servers and still have to wait for the power to run them. International Energy Agency analysis of AI and energy.

By 2031, access to electricity will play a bigger role in where companies open data centers and how fast they bring new computing capacity online. More efficient models could cut the energy needed for a single task. Total demand could still rise if cheaper AI leads people and businesses to use it more often.

What to watch — How long major data centers wait for grid connections and how much planned capacity actually comes online.

8. AI search will answer more questions and handle more of what comes next

Searching for a hotel today often means opening several sites to compare prices, locations, and refund policies. By 2031, more people could ask an AI assistant to do that comparison, build a shortlist, and prepare a booking for approval. The same pattern could apply to shopping for a laptop or choosing a subscription.

Early search data suggests why that matters for publishers. In a study of U.S. Google browsing in March 2025, Pew Research Center found that users clicked a traditional search result on 8% of visits when an AI summary appeared, compared with 15% when one did not. If that pattern continues, pages built around simple answers could see fewer visits. Firsthand reviews, original research, and information readers want to verify can still give people a reason to click. Pew Research Center’s study of Google AI summaries.

What to watch — How often AI search results lead to source visits, purchases, or bookings.

9. AI robots will spread faster in warehouses than in homes

Warehouses and factories give robots repeatable work and relatively predictable surroundings. A robot might move bins along marked routes, sort packages, or inspect items as they pass on a conveyor. A home robot has to deal with changing layouts, fragile objects, pets, and people moving through the same room.

The International Federation of Robotics reported that transportation and logistics was the largest category for professional service robots in 2025, with about 117,500 units sold. That figure covers the broader robotics category. By 2031, the stronger test for AI-enabled robots will be whether businesses pay to expand deployments after they see the machines work through busy shifts and unexpected conditions. International Federation of Robotics’ 2026 service robot report summary.

What to watch — How many customers expand an AI robot deployment from one pilot site to multiple warehouses or factories.

ALSO READ: U.S. Humanoid Robot Production Could Reach 400,000 a Year by 2030, J.P. Morgan Says

What could slow these AI predictions down?

What could slow these AI predictions down?

Errors in everyday work. An agent that handles most easy cases can still create more work if employees have to review every decision. What matters is how often it fails in real conditions, how quickly people catch those failures, and how much they cost to fix.

The cost of deploying AI at scale. Companies pay for models, computing power, data integration, monitoring, and staff time. If those costs exceed the savings from a tool, a pilot can work as expected and still stall before a wider rollout.

People’s willingness to hand over decisions. A traveler may welcome a hotel shortlist and still want to check the total price before booking. A business may trust an AI agent to prepare payments and keep a person in charge of releasing them. How much control users delegate will shape how far AI agents spread.

What will AI look like in five years?

By 2031, a useful test will be what happens when someone asks AI to do a familiar task. Can it find the right information, complete the routine steps, and hand the work back when a decision needs judgment?

The answer will show how far AI has moved into everyday life, whether the task is booking a room, resolving a refund, or moving a pallet across a warehouse.