Runs behind your software
Connects over REST to your e-commerce, CRM, ERP, SaaS or mobile application.
Ai Thinks API Platform
Integrate AI into your product instead of building it.
Text analysis, customer review scoring, content moderation, summarisation, audio transcription and smart search connect to the systems you already run through one API layer. No model of your own to build, train or host.
Already have an account? Sign in straight to the panel.
Text and reviewsAudioSmart searchContent moderation
Content from your productE-commerce
“The quality is far better than I expected and delivery was quick. Highly recommended.”
Rated 5 stars
Ai Thinks analysis
What comes back
- SentimentPositive
- Profanity and advertisingNone
- Matches the ratingConsistent
Publishing decisionPUBLISH
You define the criteria
Not a fixed black box: you set your own evaluation and critical criteria.
It returns a decision, not just data
The analysis becomes an operational call such as publish or do not publish.
How it works
Send the data, get a structured result back.
On the API side Ai Thinks never talks to the end user. It is the AI analysis and decision layer running behind other software: your system sends the data, the platform analyses it, and the result comes back in the same flow.
- 01
Your system sends the request
Your e-commerce, CRM, ERP, SaaS or mobile app passes the content to be analysed to the API.
- 02
Ai Thinks runs the analysis
The content is scored against the evaluation criteria, custom filters and weightings that you defined.
- 03
The result returns to your system
The structured output and the publishing decision come back in the same call or over a webhook.
Example: an e-commerce review
One request, five separate checks
When a customer review reaches the platform, every check you defined runs inside the same analysis and resolves into a single decision.
Customer review
“Delivery was very quick and the product matches the description. Great value for the price.”
Rated 5 stars
- SentimentPositive
- Contains advertising?No
- Contains profanity?No
- Consistent with the rating?Consistent
- Spelling and grammarCorrected
Publishing decisionPUBLISH
The decision comes from the rules you set. This review clears every critical criterion and scores above your threshold, so it goes live; had one criterion come back negative, or had the rating contradicted the review, the same request would have returned a hold decision instead.
AI analysis services
Ready-made analysis APIs for text, audio and search.
Every service can be called on its own, chained to another, and authorised separately for each company.
E-commerce review analysis
Scores customer reviews against evaluation criteria, critical criteria, the star rating and language quality — including whether the rating and the review actually agree.
Outputs
SentimentRating consistencyCritical criteriaCorrected text
Hotel review analysis
Evaluates guest reviews against quality criteria, competitor mentions, abusive language and content that may belong to a different property, then returns a publishing decision.
Outputs
Publish / holdCompetitor checkWrong-property contentCritical criteria
Review summarisation
Evaluates large volumes of customer reviews together and produces either a single overall summary or a summary broken down by category.
Outputs
Overall summaryCategory summaryCategory sentimentFocus points
Profanity analysis
Detects profanity and inappropriate language in text, and works as a moderation layer in any system where users produce content.
Outputs
swearing: true / falseInstant responseModeration layer
Advertising / referral analysis
Flags whether a review or a piece of text promotes another brand, advertises an external site, or steers the reader somewhere else.
Outputs
Competitor brandExternal referralHidden advertising
Political content analysis
Checks whether the text carries political wording or political context, and returns it as a separate signal for systems that need moderation.
Outputs
Political wordingPolitical context
Audio transcript analysis
Turns an audio recording into text: separates the speakers, splits the recording into segments, and returns either timestamped or plain-text output.
Outputs
Speaker separationTimestampsSegment listOnline / offline
Smart search
Turns a request written in natural language into the structured search parameters your own system can execute directly.
Outputs
Location & datesPrice & score rangeGuest detailsCategory & type
Which API is opened to which company is authorised separately from the panel, so a company can only call the services defined for it.
Review summarisation
Reduce thousands of reviews to a summary you can read.
Unlike single-review analysis, the summarisation service evaluates many customer reviews together and surfaces the opinions that keep repeating.
Single overall summary
Reduces every review to one piece of evaluation text.
Categorised summary
Produces a separate summary and sentiment under each heading you choose.
5,000customer reviews
Ai Thinks summarisation
You decide which topics get analysed
Product qualityPriceDeliveryPackagingCustomer service
summary per categorySample output
Product quality
MixedMost users are happy with the quality of the product, though durability draws a recurring set of complaints.
Delivery
PositiveThe majority of reviews speak positively about how quickly the order arrived.
Packaging
NegativeDamaged boxes and insufficient packaging come up repeatedly among a specific group of users.
Summaries come back in the language you need
For hotel groups, marketplaces, e-commerce companies and global SaaS applications working across borders, the output language is set with the request.
TurkishEnglishGermanFrenchSpanishItalianRussianArabicChinese
Decision engine
An AI layer that decides by your own rules.
This is what separates Ai Thinks from off-the-shelf AI APIs: you define the analysis criteria, the critical criteria and the weightings, and the platform returns the outcome as an operational decision rather than raw data.
Custom filters
A company defines its own analysis criteria, which is how the same AI infrastructure adapts to the business rules of very different industries.
Filter name
Competitor brand check
Evaluation criteria
- Is a competitor brand mentioned?
- Is a competitor product recommended?
- Does it steer the reader to an external site?
Critical criteriaIf any criterion marked as critical comes back negative, the content is held back regardless of the total score.
Weighted scoring
Analyses do not have to work as a simple yes/no. You assign a weight to each criterion and let the decision follow a minimum average score.
- Product quality%40
- Customer satisfaction%25
- Delivery%20
- Packaging%15
Minimum average weighted score60 / 100
Publishing rules
The company decides which of these rules apply.
- Hold back if a critical criterion is negative
- Hold back if it contains profanity
- Hold back if it mentions a competitor brand
- Hold back if the overall score is neutral
- Hold back if the weighted score is below the threshold
PUBLISHUNPUBLISH
That turns Ai Thinks from something that merely analyses into an API layer that produces decisions from your business rules.
Audio intelligence
Turn recordings into speaker-separated transcripts.
Call centre recordings, conversations and meeting audio become text: speakers are separated and every segment is marked with a start and end time.
Call centre recordingAudio transcript analysis · English
- Speaker 100:00 – 00:12Hello, the order I placed last week still has not reached me.
- Speaker 200:13 – 00:37Let me check that right away — could I take your order number?
- Speaker 100:38 – 00:46SP-284193.
Transcript options
A transcript can be chained into the other analysis services
- Phone call
- Transcript
- Summary
- Sentiment analysis
- Complaint topics
This chain is one possible way of composing the existing services; each step is a separate API call.
Smart search
Turn a request in plain language into a structured query.
Takes what a user expressed in a sentence and converts it into search parameters your own system can execute directly.
What the user typed
“Find a family hotel in Antalya between 15 and 20 August for two adults and one child, within a certain price range.”
Generated search parameters
- CityAntalya
- Check-in15 August 2026
- Check-out20 August 2026
- Adults2
- Children1
- Price range€230 – €400
- Hotel categoryFamily hotel
- Meal typeAll inclusive
Supported search criteria
CityCountryDistrictStateGeo IDLatitude / longitudePrice rangeScore rangeCheck-in dateCheck-out dateNumber of adultsChildren’s dates of birthHotel featuresHotel categoriesHotel typesMeal type
The parameter model is mapped to hotel and travel data; the same structure can be extended to your own data model.
API management
Manage access per company.
The APIs are never left openly accessible. Which company may use which service, and where the analysis result should go, are defined from a single panel.
Company-level API permissions
Not every company needs every API; permissions are granted and revoked per company.
Company A
- E-commerce review analysis
- Profanity analysis
- Hotel review analysis
- Smart search
Company B
- Hotel review analysis
- Smart search
- Audio transcript
- Profanity analysis
Asynchronous results over webhooks
You define the address the result should return to, so long-running analyses do not force the external system to keep polling the API.
Webhook configuration
- Webhook name
- Target address
- Security headers
- Start the analysis
- Ai Thinks processes it
- Analysis completed
- Webhook fires
- Result in your system
The platform calls you once the result is ready — no open connection to hold.
Solutions on this platform
Ready-made solutions running on the API Platform
You do not have to build on the services from scratch. Start from scenarios already proven in the field and adapt them to your own business rules.
Ai Thinks API Platform
Let us connect AI to your product.
We review your data sources, your analysis criteria and your integration scenario together, then set out a pilot plan with a test token.