Skip to content

Early access

The rates, offers and answers banks show only to people who apply.

Every step of account opening up to the identity check, the rates and switching offers shown along the way, every email and text after, and what chat tells a saver. At your bank and your competitors, each dated with its screenshot or transcript.

Onboarding a limited number of teams. Name the AI agent to start with.

We keep what you tell us to set up your access and email you about Obsession. Privacy notice

  • Kittiwake Bank, a savings bank
  • 3 competitors
  • Sign-up, login and chat

  1. What you name

    Your bank and 3 competitors

    • Account opening, up to the identity check
    • Rate pages and the application
    • Chat and the help centre
    • The customer profiles to ask as

    A new saver and a switcher

  2. What comes back

    Every fact, dated

    • Each step, up to the identity check
    • Every email and text after
    • Application rates beside public ones
    • Chat answers, word for word

    Each with its screenshot or transcript

  3. Where it lands

    Where your team works

    • Search: one bank
    • Insights: a question across many
    • A webhook, Clay or the API

    Nothing to install

Name the banks and who asks. Everything after is kept.

Kittiwake Bank, a savings bank, follows its own channels and 3 competitors’.

Set it up

Which banks, as which customers, and what each answer is checked against.

  • Your bank and any competitor
  • New saver · UK
  • Switcher · UK
  • Says it’s AI to a person

  1. Name the banks

    Your own bank and each competitor, on the web or in the app.

    ExampleKittiwake names itself and 3 competitors: Dunmere, Sedgefield and Quaystone.

  2. Pick who asks

    The customer profiles and country to ask as. A new saver and a switcher are often offered different rates.

    ExampleA new saver in the UK, and someone switching their savings from another bank.

  3. Say what to check against

    Your rate page and your terms. An answer that disagrees with them becomes a finding, with both side by side.

    ExampleThe rate, fixed or variable, the notice period and the bonus end date.

Obsession starts the applications

Up to the identity check, never past it. Each keeps its inbox and phone number, so every email, text and offer after arrives as it happens.

  • Its own inbox
  • Its own phone number
  • Its own browser
  • Says it’s AI

  1. Account opening, step by step

    Every form up to the identity check, what that check asks for, and the emails after it, each with the time it arrived.

    Example1 Nov, 14:22 at Dunmere: the application stops at the identity check, which asks for a passport photo and a selfie. At 48 hours: “Finish your application”.

  2. Application rates beside public ones

    The rate and bonus the application shows before the identity check, beside the ones on the public page.

    ExampleSedgefield: 4.8% on the public page, 4.5% on the application’s summary page.

  3. What chat tells a customer

    The same questions in chat and support, each answer kept word for word and checked against the terms.

    ExampleTue 07:04, Kittiwake web chat: “Yes, 4.6% is fixed for 12 months.” The rate page says variable.

Rate tables show the public rate. The application can show another.

Comparison sites stop at the public page, and panels sell a fixed list of banks behind a client login. The rest is found by staff starting applications by hand, a few banks at a time.

  • Rates in the application

    A rate or bonus on the summary page that differs from the public one.

  • Offers after the form

    A switching bonus shown only once the form is filled in.

  • Wrong answers in chat

    A bot that calls a variable rate fixed, or gets a notice period wrong.

  • Messages that never come

    The reminder to finish an application that never went out.

Each finding is dated, with the screenshot, email or transcript behind it, kept for as long as you need it.

What product, pricing and risk teams do with it.

  • Price against what applicants see

    Not the rate on a comparison table.

    ExampleAsked “Which banks show a lower rate in the application than on their rate page?”, Insights answered 3 of 5, by up to 0.3 points.

  • Catch a wrong answer the day it starts

    Not when the complaints arrive.

    ExampleThe fixed rate answer reached the savings team’s webhook at 07:05 on Tuesday, the morning after a bot update.

  • Show what customers were told

    A dated record of every answer, for your own monitoring and your board.

    ExampleOctober: 40 questions a day in 3 channels, 4 wrong answers, each fixed within a day.

  • Match a competitor’s offer, or don’t

    Switching bonuses and rate changes, the day they appear.

    ExampleQuaystone raised its easy access rate to 4.75% on Thursday. Its chat quoted 4.5% until Friday.

Guides for this job

Step-by-step guides behind this use case, each with what a pass looks like.

What banks and fintechs ask.

Obsession records what banks and fintechs show only to the people who apply: every step of account opening up to the identity check, the rates and offers shown along the way, the emails and texts after it, and what chat tells a saver. Each fact is dated, with its screenshot, email or transcript, for your own bank and your competitors. Look up one bank, or ask a question across many.

From real applications. Obsession’s agents start applications, sign up where a bank allows it and ask chat the way any customer can. They ask as the customer profiles and countries you choose, and keep each inbox and phone number for months. They stop at the identity check and never invent identity details. They never sign in to an account that isn’t their own, and say they’re an AI when they talk to a person.

In the UK, the Consumer Duty asks firms to check that what they tell customers supports their understanding. Obsession gives you a dated record of what your channels said, answer by answer. It isn’t compliance advice, and what you report stays your call.

Only your team. Anyone can see that a bank is covered, and since when, but never who asked. Your own questions and their transcripts stay on an account that works only for you.

Each application keeps its own inbox and phone number, so emails and texts arrive as they happen. Rate pages and chat questions are checked every morning in this example, again after every bot update, or on any rhythm you set.

Which banks should we look up first?

Early access includes API and MCP access, lookups of every company already covered, and up to 5 companies added for you, set up with the founders.

Onboarding a limited number of teams. Name the AI agent to start with.

We keep what you tell us to set up your access and email you about Obsession. Privacy notice