Connor
Connor reviews financial promotions before they are published and monitors them once they are live. A marketing or compliance team submits an advert, social post, video, PDF or web page, and Connor checks it against the rules for the firm's home market, each market it targets and the products it promotes. Each finding shows the flagged text, the rule it breaks, a severity and a suggested fix. A named approver at the firm makes every decision.
It is built for brokers, investment platforms, crypto firms, lenders, banks and insurers, with rulebooks covering 41 markets. A separate monitoring product checks affiliates' and influencers' public accounts, hashtags and the firm's own sites on a schedule, and reviews the relevant posts the same way. An assistant, Ask Connor, answers rule questions inside Slack, Microsoft Teams, Claude and ChatGPT and quotes each provision it relies on.
Connor publishes its data processing agreement and a full subprocessor list, with Anthropic and OpenAI named for the AI review. It names no customers and no security certification. Pricing is a monthly platform fee plus credits for each promotion checked, after a free first week. Connor is the trading name of Zindex Limited, registered in England and Wales.
Capability Axes
Capability grades
15 of 15 axes rated · 11 graded A or B
The review is the models. Connor reads each advert, video, PDF or web page, checks it against the rules that apply to that firm, market and product, and writes up each breach with the rule it breaks and a fix. A model also transcribes speech in videos, and another screens monitored posts for relevance before they are reviewed. The pricing says the same thing: credits are spent on each promotion checked, while approvals, comments and reports cost none. Without the models, what remains is an approval queue and a regulation library, the parts a firm can already get from a workflow tool.
Connor approves nothing itself. A named approver at the firm decides every version, and the terms make the firm solely responsible for whether a promotion is approved, amended, withdrawn or published. Each team builds its own workflow, such as submission, AI review, compliance and legal, can require one approver or all of them, and can stop people approving their own submissions.
A workflow change applies only to the next submission, a new version restarts review, and the history of who decided is kept and can be exported for audit. On the monitoring side the firm sets the relevance bar a post must clear to be reviewed. Posts below it are stored but not reviewed, so where that bar sits decides what a person never sees.
Every finding can be checked against its source. It shows the flagged text, the rule it breaks, a severity rank and a suggested fix, and each rule links to the regulator's own text, marked as binding rule or guidance. Only the rules that apply to the firm's markets and products run, and a firm can switch rules off, change their severity or add its own. Connor publishes no accuracy figure, validation method or error rate.
Its terms state that findings may be incomplete or inaccurate and that an incomplete review must not be treated as a pass, so the figure that matters most, how often a breach goes unflagged, is left for the firm to establish in its trial.
Connor names no customer and publishes no case study, deployment count or measured result. The approval rate and decision times shown in its reporting screens belong to a product demo, and the post counts on its monitoring dashboard are labeled as demo figures. Its Ask Connor page shows an answer it describes as shortened from a real one of 18 October 2026, which shows the product in use without saying by whom. A firm evaluating it has no public reference in its own segment or market to check against.
The data processing agreement commits Connor not to use customer personal data to train or fine tune any model shared with other customers, or to build a shared evaluation corpus from it. Its security annex says the AI providers it uses are bound by terms that prohibit training on customer data, and each request is authorized against the customer and the user's role. The commitment is written about personal data.
It does not say whether one firm's promotions, house rules and approval decisions shape the checks another firm receives, which is what a compliance team will want to know about its own interpretations. How long the model providers keep prompts and outputs is not stated.
The data processing agreement, version 1.2 of 2 November 2026, is published as part of the terms. It applies United Kingdom and EU data protection law, lists fourteen subprocessors with their roles and locations, and sets exit terms of thirty days to export after termination and deletion within a further thirty. It also names whose data the service handles, including the presenters, endorsers and creators in submitted or monitored promotions, with their social handles and comments.
The duty to give those people any required notice sits with the customer, so a firm monitoring influencers carries the lawful basis for doing so. Backup periods and transfer safeguards are set in a per customer deployment schedule that is not public.
Connor names no security certification or independent audit report. The security annex in its data processing agreement describes encryption in transit and at rest through its hosting and database providers, unique user accounts with role based access inside each customer workspace, and incident notice without undue delay. The annex calls itself a contractual baseline rather than a certification of completed testing, which is an accurate account of what it is. Multifactor sign in, penetration testing and a trust center are not described, and these are the first items an enterprise security review will request.
The terms state the correct position for a technology supplier. The service is described as a compliance aid, not legal or regulatory advice, and not approval by Connor under the Financial Services and Markets Act 2000, so the approval duty stays with the firm. Its rule engagement is specific.
Rulebooks cite the Financial Conduct Authority's conduct of business, consumer credit, banking, insurance and mortgage sourcebooks and MiCA, and the regulators named include CySEC, ASIC, FINRA, the SEC, the Central Bank of Ireland and VARA. Connor names no license, regulatory program or register entry, which is the expected position for a supplier of this kind.
Connor publishes no governance framework, fairness position or testing program for its models. The product judges marketing rather than people, so the bias surface is modest, with two exceptions. Monitoring reviews posts by named influencers and affiliates, and speech in videos is transcribed by a model whose accuracy can vary with accent and recording quality. Review depth is the other question. With rulebooks across 41 markets, coverage is likely deeper in some markets than others, and nothing published says where it is thinner.
The terms put the outcome on the firm. Connor does not warrant that the service will find every non compliant item, states that findings may be incomplete or inaccurate, and makes the firm solely responsible for whether a promotion is approved, amended or published. It keeps a duty to perform with reasonable skill and care, caps its liability at the greater of £10,000 and a year's fees, and gives the customer no indemnity.
What stands behind each output is an audit trail of every version, finding, comment and decision, exportable as a PDF for auditors or regulators. A creator whose post is flagged has no route to contest the finding, only the fix the firm may choose to send.
The full subprocessor list is public: fourteen providers, each with its role and location. Anthropic and OpenAI carry out the AI review, OpenAI also transcribes audio, and TypeSafe AI screens monitored posts for relevance. Vercel, Supabase and Railway host the application, files, database and background work. Clerk handles sign in and Browserbase renders web pages.
Apify, Decodo, the X interface and Google's YouTube interface retrieve posts and videos, Resend sends notifications and Slack carries internal alerts. The list notes that each customer's deployment schedule decides which providers apply to it, and model versions are not named.
Promotions arrive as uploads, pasted copy, links to social posts or web pages, which Connor renders as a visitor in each market would see them. Ask Connor works inside Slack, Microsoft Teams, Claude and ChatGPT. The company also states an interface for a firm's own systems but links no public documentation for it. Monitoring reads Instagram, TikTok, YouTube, Facebook, LinkedIn and X, following company pages rather than personal profiles on LinkedIn. Connor names no connection to the asset libraries, content systems or project tools where marketing work is made, so material reaches review only when someone submits it.
Hosting is cloud only, and the locations are published. The data processing agreement places application hosting in Dublin, file storage in London, the database in Ireland and background processing in Amsterdam. The AI providers are listed as processing in the United States and other regions, with transfer safeguards set out in each customer's deployment schedule rather than on the public page. The pages offer no choice of region and no private or on premise deployment. For a firm with a residency requirement, the location that matters is where the model calls run, since that is where each promotion is read.
The charging basis is published and the rates are not. A customer pays a monthly platform fee plus credits for each promotion checked, both scaled to usage, and approvals, comments and reports use no credits. The first week is a free trial, and onboarding, which builds the firm's rulebooks and calibrates the checks, takes about a week. The terms add that paid subscriptions run for twelve months and renew automatically unless either side gives sixty days' notice. A firm can see what drives cost before a call, but not the fee or the price of a credit.
Seven segments each have their own page: CFD and forex brokers, investment platforms and funds, crypto firms, consumer credit lenders and brokers, mortgage lenders and brokers, banks and savings providers, and insurers and insurance brokers. The pages carry segment rules rather than shared copy. The crypto page works through the cryptoasset risk warning, incentive and cooling off rules in the Financial Conduct Authority's conduct of business sourcebook.
The mortgage page covers representative examples, secured lending warnings and green mortgage claims under its mortgage sourcebook. Rulebooks span 41 markets, from the United Kingdom and European Union to the United States, the Gulf states, Hong Kong, Singapore, Japan, India and South Africa. The segment pages illustrate British rules only, so a firm elsewhere will want to see its own market's rulebook during the trial.
Alternatives to Connor
Other products that do the same kind of work as Connor, and what each one does.
Finspector reviews financial promotions for regulatory risk and then keeps watching them once they are live.
Adclear reviews financial promotions for regulatory compliance before they go live and monitors them once published, replacing a manual approval loop that ran in days with feedback in seconds.
bethebrand sells marketing compliance software to UK financial services firms.
PerformLine runs marketing compliance oversight across every channel a regulated brand speaks through, covering both halves of the lifecycle in one platform: automated review and scoring of assets before publication, then continuous discovery and monitoring after launch across web pages…
Saifr sells AI agents that read financial marketing and communications and flag regulatory risk before anything is published.
Luthor reviews regulated marketing before it goes live and describes itself as an artificial intelligence native compliance firm rather than a pure software vendor.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
Not published. A monthly platform fee plus credits for each promotion checked, after a free first week.
|
Charged as a monthly platform fee plus credits spent on each promotion checked, both scaled to usage. Approvals, comments and reports use no credits. Paid subscriptions run for twelve months and renew automatically unless either side gives sixty days' notice before the term ends. | Data protection terms are not tiered. The data processing agreement forms part of the terms for every customer. | No implementation fee is published. The first week is a free trial, and onboarding, which builds the firm's rulebooks and calibrates the checks, takes about a week. Before an order form is signed the service runs as an evaluation with no fees unless agreed in writing. | Vendor Published |
Connor charges a monthly platform fee plus credits for each promotion checked, and quotes both amounts rather than publishing them. Approvals, comments and reports use no credits.
Invoices are payable within 15 days and fees exclude VAT. Fee changes at renewal need 60 days' notice, and a pilot ends on its end date without converting to a paid subscription on its own.