# Back-office & Internal Tools Adoption | Adoption Layer

> Guide your teams inside bespoke back-offices and internal business apps: the AI reads the live screen, with no documentation or selectors to maintain.

## Your bespoke back-offices, finally usable without the colleague who "knows".

Dense screens, unlabeled buttons, rules known to three people, documentation dated from V2: your internal applications run the business and nobody can use them without help. Adoption Layer reads the live screen and guides every employee, even on a tool you've never documented.

## The know-how lives in three people's heads, not in the tool

Your back-offices were built for the business, not for usability. Labels are codes, validations arrive with no explanation, and edge cases get resolved by asking the veteran. Every departure takes part of the manual with it, every arrival costs weeks of shadowing, and the product team has no bandwidth to rebuild or to document.

> "A case handler needs to close a contract in the in-house back-office: 'The Close button is greyed out, why?' Adoption Layer reads the screen, spots an amendment still in Draft status in the History tab, highlights the relevant row, and states the action to take before closing, in line with the internal procedure."

## Guidance that works on a tool nobody has documented

- **Live reading of bespoke interfaces**: Dense tables, nested tabs, unlabeled buttons, dynamically generated forms: the AI interprets the DOM as it renders, with no prior mapping and no standard components required.
- **Capturing tacit knowledge**: The rules only the veterans know are written in plain language in the dashboard and surfaced on the relevant screen. No wiki to maintain, no screenshots to redo.
- **Onboarding without shadowing**: A new hire handles their first cases guided step by step, without tying up a colleague for three weeks.
- **Friction reports for the product team**: Recurring questions, screens where users get stuck, and misunderstood labels are consolidated to help you prioritize improvements.

## Business workflows covered

- Processing and closing cases in operational back-offices
- Entering and approving transactions in internal admin tools
- Onboarding new employees on company-specific tools
- Migrating from a legacy back-office to its new version

## Why classic DAPs don't fit your internal tools

WalkMe and Lemon Learning are built for standard SaaS where someone has already recorded flows. On an internal application, you first have to map every screen, place selectors on components with no stable identifiers, then redo it all at every deployment. Adoption Layer stores no selectors: it reads the DOM displayed at the moment of the question and guides the user on the real screen, even if the interface changed yesterday.

## Setup and privacy

Add one line of script to your application (or deploy the Chrome extension via Intune or Jamf if you'd rather not touch the code), allow-list your internal domains, and write your business rules in the dashboard. Case data, identifiers, amounts, and contact details are redacted in the browser before any inference. No cookies, no screenshots.

## Frequently asked questions

### Our application has no documentation. Do we have to write it all before starting?

No. The AI already guides on the screen structure (buttons, fields, tabs). You then add your business rules as you go, starting from the questions users actually ask.

### Does it work on old or non-standard interfaces (HTML tables, iframes, in-house frameworks)?

Yes. The extension reads the DOM rendered by the browser regardless of framework. Same-origin iframes are covered; non-HTML content (Java, Citrix, thick clients) is not.

### Script or extension: which should we choose?

One line of script if you control the application's deployment and want to cover every browser; the Chrome extension via MDM if you can't or won't touch the code. Both deliver the same guidance.

### Is the data in the cases being processed protected?

Yes. Redaction happens 100% client-side: field values, identifiers, amounts, and contact details are replaced with tokens before inference. You can also exclude entire areas of the screen.