Every AI use case, mapped to the business.

32 use cases owners are being sold this year, sorted by where they act in a business of 5 to 100 people. Only 3 are proven in production.

Status
Pick a room to see its use cases
How we rate

Four statuses, one rule: rate the evidence, not the pitch.

Each entry gets a status for how proven it is in businesses of 5 to 100 people, and a separate rating for how strong the evidence behind that status is.

Evidence strength

StrongSeveral independent studies or large independent surveys.
ModerateAt least one independent study or credible survey, plus consistent case evidence.
ThinA handful of independent data points.
Vendor claimsOnly what sellers publish about their own products.

How an entry is made

Each entry is drafted from public sources: academic studies, regulators, independent surveys and reputable reporting. Every figure carries a numbered citation you can open. Anything a vendor says about its own product is labelled as a vendor claim.

Review and updates

Every source in this edition was opened and checked against the text in October 2026. Each entry is then signed off by a named operator and re-checked when the evidence moves. Until that sign-off, the entry is marked as a draft at the top of the page.

The full guide

All use cases at a glance

Field GuideBrand, marketing & customersAI review responses
In pilotsBrand, marketing & customers

AI review responses

Also called AI replies to Google reviews, automated review replies, AI reputation management, review reply generator

EvidenceThin
ClassGenerative

Replying to reviews is linked to better ratings, but evidence on AI-written replies is thin and disclosure hurts. Research, rules and costs.

ReviewDraft, awaiting operator review
Drafted27 October 2026
Fact-checkedEvery source opened, October 2026
Key takeaway

Replying to reviews matters: hotels that started responding gained 0.12 stars and 12% more reviews⁠, and for independent restaurants one extra star is linked to 5 to 9% more revenue⁠. Whether AI should write the replies is less clear. ChatGPT drafts ran far longer than managers' replies and apologised far more often⁠, and customers told a reply was AI-written were less satisfied⁠. Let AI draft and a person send.

At a glance

StatusIn pilots
What it doesDrafts replies to Google, Facebook, marketplace and booking-site reviews in a consistent tone, for a person to edit and post.
Typical costReplying on a Google Business Profile is free⁠. AI drafting tools vary by vendor; dedicated review-management tools are usually quote-based.
Data it needsYour reviews across platforms, current policies and offers, and a list of what staff may and may not promise in public.
Setup effortLow: copy, draft, edit, post. Higher if you connect review platforms to a tool.
EvidenceThin. Good independent evidence that replying helps; thin and mixed evidence on AI-written replies specifically.

What are AI review responses?

AI review responses are replies to customer reviews drafted by a language model. An owner pastes a review into a chatbot, or a review-management tool pulls reviews from Google, Facebook or a marketplace, and the AI writes a reply in the business's voice. A person should then edit and post it.

The appeal for a busy owner is obvious. Reviews arrive at awkward times, negative ones sting, and a backlog of unanswered reviews looks careless.

This entry is about replies. It is not about using AI to fake reviews, which is a different practice and is banned in the US and UK. The US Federal Trade Commission's 2024 rule bans fake reviews, including AI-generated ones⁠, and in 2024 the FTC brought a complaint against an AI writing tool whose review feature let users produce reviews in bulk⁠.

How does it work?

The model reads the review and produces a reply that thanks, apologises or explains. Better set-ups also give it the business's policies and a list of things it must never say, then queue drafts for approval.

The output has a recognisable style. A study of 115 complaints at Hong Kong luxury hotels compared ChatGPT drafts with managers' actual replies⁠:

  • AI replies averaged 211 words against the managers' 130, and met a 130-word limit in only 2 of 115 cases.
  • The AI's replies used the word 'apologise' 213 times against the managers' 19.
  • It mentioned complimentary items 21 times against the managers' once.
  • Managers more often moved detail to private channels and invited guests back.

That matters for a small firm. An over-apologetic reply can read as an admission of fault, and an unprompted freebie is a promise someone has to honour.

What it needs to work

Good AI replies depend on knowing your policies, your reply style and what you may promise in public; judging them depends on review data tracked over time.

  1. Access to every review platform you use, typically Google Business Profile, Facebook, Takealot or booking sites, with one person accountable for replies
  2. A short reply policy: tone, sign-off name, what may be offered in public (refunds, vouchers) and what must move to a private channel
  3. A list of things the AI must never mention in public, such as health details, order numbers or staff names
  4. At least 12 months of review history in a spreadsheet: date, platform, star rating, and whether and when you replied, to set a baseline
  5. Current policies and offers, so drafts do not promise something you have stopped doing

Most use cases in this guide succeed or fail here, on the data underneath them, before any AI is involved.

What does the evidence show?

The case for replying at all is reasonably strong. Using differences in reply rates across review platforms, researchers found hotels that began responding gained 0.12 stars and 12% more reviews, and received fewer negative reviews⁠. Ratings matter commercially: for independent restaurants on Yelp, one extra star is linked to 5 to 9% more revenue, with no effect for chains⁠.

The case for AI writing those replies is thinner:

  • In a study of 32,129 hotel reviews, managers' replies that closely resembled default ChatGPT output were linked to fewer helpfulness votes, though resemblance to more creative AI settings helped on positive reviews⁠. Readers never saw actual AI replies in that study⁠.
  • In health-care experiments, consumers generally could not tell AI replies from human ones, but when told a reply was AI-written they were less satisfied and less likely to consider the provider⁠.

We found no independent study showing AI-drafted replies improve ratings or revenue for small businesses.

AI-drafted replies vs replies written by the owner

Compared onAI-drafted, person-approvedOwner-written
SpeedSeconds per draftMinutes per reply, often delayed
Length and toneLonger and more apologetic⁠Shorter; more often invites the customer back⁠
Reaction if disclosedLower customer satisfaction⁠No disclosure issue
Main riskUnintended offers and repeated personal details⁠Defensive or angry replies
Best forHigh volumes of positive reviewsNegative or sensitive reviews

Status and evidence

StatusIn pilots
Evidence strengthThin

We rate this pilots. Independent studies show that replying to reviews can lift ratings⁠ and that ratings affect independent firms' revenue⁠. Evidence on AI-written replies is thin and mixed: one observational study⁠, one linguistic comparison⁠ and one set of health-care experiments showing a disclosure penalty⁠.

Where does it go wrong?

  • Promises you did not mean. AI drafts offered complimentary items far more often than managers did⁠. Every public offer is a commitment.
  • Sensitive details repeated in public. In the health-care study, ChatGPT sometimes repeated a reviewer's medical information and implied actions that were not under way⁠.
  • Getting found out. Disclosure lowered satisfaction, and a label added by the review platform hurt more than the business disclosing its own AI use⁠.
  • Arguing with reviewers. The FTC rule allows public replies to negative reviews but bans knowingly false accusations against reviewers and threats used to get reviews removed⁠.
  • Sliding into fake reviews. Using AI to fake reviews, including reviews from people who do not exist, is banned in the US⁠. In the UK, fake and concealed incentivised reviews became banned practices in April 2025, with fines of up to 10% of global turnover, and the CMA opened five investigations in March 2026⁠.
  • Every reply sounds the same. Replies resembling default AI output drew fewer helpfulness votes⁠.

Numbers it should move

Response timeHow long a customer waits for a first reply.
Repeat customersThe share of customers who come back and buy again.

Ask the demo business

Northgate Supply is a fictional wholesale distributor of catering and packaging supplies: 28 staff, R46.6m in sales over the last year. Its data is generated for this demo, to 30 September 2026.

Fictional data
How fast do we answer customer queries?
2.2 hoursmedian first reply in September

Half of September's 180 customer queries got a first reply within 2.2 hours, but 3 waited more than a working day. December was the slowest month, at 4.7 hours.

OctNovDecJanFebMarAprMayJunJulAugSep2.2 h
How this was worked out

Median hours from a customer query arriving to the first reply, per month, across email and WhatsApp.

Answered in 1 msScanned 1 959 rowsFrom customer queries
Want this on your own numbers, in the AI tools you already use?Book a demo

Where the data usually lives

Google Business ProfileTakealot Seller PortalShopifyWooCommerceEmail inboxesExcel and Google Sheets

Where the records this use case needs are usually kept in a small business. It is not a list of recommended tools.

Frequently asked questions

How much does it cost to use AI to reply to Google reviews?

Creating a Google Business Profile and replying to reviews on it is free⁠. AI drafting tools vary by vendor, and dedicated review-management tools are usually quote-based. The real cost is the time it takes to check each draft, so it does not promise refunds or freebies you did not intend; AI drafts offered complimentary items far more often than managers did⁠.

Do I need a developer to automate review replies?

No. Copying a review into an AI tool, editing the draft and posting it needs no technical help, and connecting review platforms to a dedicated tool is usually done in that tool's settings. Avoid fully automatic posting: the evidence on AI replies is mixed⁠, so a person should approve every public response.

Is it legal to use AI to reply to customer reviews?

Replying with AI help is not banned in the US or UK. What is banned is faking reviews, including with AI⁠, and using threats or knowingly false accusations against reviewers⁠. In the UK, the CMA can fine up to 10% of global turnover for banned review practices⁠. This is general information, not legal advice.

Should I tell customers my review replies are written by AI?

If you use AI, say so yourself and have a person edit and sign every reply; that is what the researchers recommended⁠. In health-care experiments, people told a reply was AI-written were less satisfied, but the drop was smaller when the business disclosed its own AI use than when the platform added a label⁠.

What do POPIA and UK GDPR mean for replying to reviews with AI?

Reviews contain personal information, such as names and sometimes health or order details. POPIA requires security safeguards and a written contract with any operator, such as an AI vendor, that processes that information for you⁠. The UK ICO's guidance on AI and data protection covers lawfulness, transparency and accuracy⁠. Never repeat personal details in a public reply. This is not legal advice.

More in brand, marketing & customers

AI ad creative generationIn pilotsEvidence: ModerateCustomer-service chatbotsIn pilotsEvidence: StrongAI-written social contentCosmeticEvidence: Thin

Sources

  1. 1
    Online reputation management: estimating the impact of management responses on consumer reviews
    Marketing Science (Proserpio and Zervas), via Boston University OpenBU, 2017Study
  2. 2
  3. 3
    Generative AI vs. Humans in Online Hotel Review Management: A Task-Technology Fit Perspective
    Tourism Management (Zhang, Xiang and Zach), via Virginia Tech VTechWorks, 2025Study
  4. 4
  5. 5
    Should Businesses Use AI to Respond to Online Reviews?
    University of Arkansas Walton College (Villanova and Mousavi, Journal of Public Policy & Marketing), 2026Study
  6. 6
  7. 7
  8. 8
    FTC Announces Crackdown on Deceptive AI Claims and Schemes
    US Federal Trade Commission, 2024Government
  9. 9
    Fake and misleading reviews: 5 businesses under CMA investigation
    UK Competition and Markets Authority, 2026Government
  10. 10
    Guidance on AI and data protection
    Information Commissioner's Office (UK), 2023Government
  11. 11
    Google Business Profile
    Google, 2026Vendor
  12. 12
    Protection of Personal Information Act 4 of 2013
    Government of South Africa, 2013Government

This entry is educational. It is not legal, financial or professional advice, and it does not recommend any vendor. Figures attributed to vendors are their own claims.

Ask your own business. Get the true number in seconds.

Veld connects your accounting, sales and operations data to the AI tools you already use, and keeps it running, so every answer comes from your numbers.

Book a demo