REWORK · Applied organisational AI
REWORK is Retoba's selected practice for organisational AI: copilots, agents and workflows designed around real work, defined responsibility and human judgement.
From a working problem to a responsible system
Why REWORK exists
Retoba believes in strong focus. Some of our most useful lessons began when trusted partners brought us a specific working problem and asked whether AI could help.
REWORK is a selected archive of those purposeful explorations. It applies Retoba's disciplines of knowledge architecture, governed generation, integration and measurement to the work happening inside organisations.
The projects taught us how professionals search, judge, create, communicate, decide and hand work over.
Operating principles
REWORK is not a catalogue of generic assistants. Each exploration starts with a real role, an accountable owner and a reason to change the way work is done.
The task, decision and expected value are defined before the technology.
Roles, permissions and moments of human review remain explicit.
Authorised sources, traceability and uncertainty are part of the system.
A useful exploration earns its next step through observed work, not novelty.
Selected work
These anonymised examples show how Retoba approaches specific working problems with AI systems designed for a clear purpose.
Newsroom · Copilot
Challenge
A newsroom needed faster research and drafting without losing its editorial standards or each journalist's individual voice.
Custom solution
Retoba built a copilot combining the publication's rules with authorised author profiles for research, drafting and editing.
Reputation · Copilot
Challenge
A PR team needed to respond quickly while working within approved facts, mandates and reputational boundaries.
Custom solution
Retoba built a governance copilot that finds relevant information, tests response options and flags issues for professional review.
Global marketing · Platform
Challenge
A global consumer technology company needed more creative variants without compromising product accuracy, brand rules or control.
Custom solution
Retoba built a governed platform connecting briefs, approved assets, generation, validation, revisions and human approval across markets.
Legal work · Copilot
Challenge
An international law firm needed faster document analysis and drafting without relying on unsupported legal answers.
Custom solution
Retoba built a copilot grounded in authorised sources. It compares approved materials, prepares traceable drafts and leaves legal judgement with the lawyer.
Leadership · Copilot
Challenge
A leadership team needed to examine complex decisions without hiding uncertainty, assumptions or missing evidence.
Custom solution
Retoba built a decision copilot that structures options, scenarios, risks and indicators before the responsible leader decides.
Sales training · Agent
Challenge
A sales organisation needed repeatable training that reflected its own products, buyer roles and real objections without depending on a trainer for every exercise.
Custom solution
Retoba built a sales training agent that adopts realistic buyer profiles, reveals information in response to good questions, adapts objections and gives structured feedback after each simulation.
Regulatory · Agent
Challenge
A nutritional supplements distributor needed to assess ingredients, labelling and health claims across EU, Slovenian and Croatian rules without relying on fragmented sources or assumptions.
Custom solution
Retoba built a regulatory agent grounded in authorised sources. It checks approved legislation and data on botanical claims, cites relevant provisions, distinguishes national exceptions and flags uncertainty for professional review.
Advertising operations · Agent
Challenge
A digital agency needed one operational entry point across advertising platforms, markets, analytics and CRM, with continuous optimisation and clear control of budgets and approvals.
Custom solution
Retoba built an agent that works across Google, YouTube, Meta, TikTok, LinkedIn, Amazon Ads and the programmatic stack. It uses analytics and CRM/CDP signals to optimise continuously, records every change and routes actions with major consequences for approval.
Finance and controlling · Copilot
Challenge
A management team needed one reliable view across fragmented finance, sales, workforce and accounting files without manually reconciling every workbook before each decision.
Custom solution
Retoba built this controlling copilot for a company with 60 employees. It validates and normalises approved data from several years, keeps its source base traceable and prepares management views of performance, budget variance, client profitability, costs and workforce indicators. Management reviews the evidence and decides.
Market intelligence · Copilot
Challenge
A specialist sports law education initiative had a detailed curriculum but needed a clear basis for decisions about demand, competitive position and commercial viability before launch.
Custom solution
Retoba developed a custom market strategy copilot that combined the supplied programme data with international market and competitor research. It structured target audiences, positioning, pricing scenarios, market entry priorities and an action plan for the next 12 months, while final strategic choices remained with the project team.
Leadership operations · Agent
Challenge
A mid-sized company needed to turn information scattered across finance, sales, CRM, project management and business documents into coordinated action. Management had to gather context, assign follow-up work and track progress across separate tools, while keeping decisions and responsibilities clear.
Custom solution
Retoba configured an agent that works across connected business tools, bringing relevant information, previous decisions and open tasks into one working context. It checks sources, prepares decision briefs, carries out authorised tasks and delegates follow-up work through those tools. Responsibilities, progress and results remain traceable across conversations, while consequential actions require management approval.
Engineering & real estate development · Agent
Challenge
An engineering and real estate development company needs to gather information from public portals and internal systems before it can assess a site, prepare a tender or move a project forward. Teams have to find property records, check planning conditions, collect drawings and compare offers, then turn the findings into documents and follow-up work.
Custom solution
Retoba developed an agent that takes a project brief and carries out this preparation. It searches government and municipal portals, retrieves information from company systems and brings together the records needed for the task.
The agent prepares project reports, offer comparisons and draft tender or application documents. It saves the material in the project workspace, updates project records and assigns follow-up tasks through connected tools. Findings link to their sources, and missing or conflicting information is highlighted for review. Specialists check technical conclusions and approve formal submissions.
Digital operations · Agent
Challenge
A growing AI company needed to evolve its website continuously without destabilising the live experience or losing past decisions.
Custom solution
Retoba configured a supervised agent that researches, writes, designs, builds, previews, verifies and documents approved changes. Publication remains an explicit owner decision.
This website is part of the evidence.
What the work taught us
Across very different roles, the same design conditions kept returning.
Useful output begins with the organisation's language, sources, standards and actual task.
The system should show what it knows, what supports the answer and where uncertainty remains.
The ability to act is designed as a bounded responsibility, not treated as a technical default.
A polished conversation is useful only when it fits the decisions, systems and handovers around it.
AI can prepare, compare, test and recommend. Responsibility remains with the person or organisation authorised to decide.
Precise language
Clear terminology sets expectations about capability, control and responsibility. REWORK uses four practical categories.
Supports a person inside a defined task. It prepares work; the person reviews and decides.
An AI agent performs a bounded sequence of actions within explicit permissions, controls and escalation points.
Connects people, knowledge, systems and approvals across a repeatable process.
Provides a shared governed environment for multiple roles, workflows and reusable capabilities.
Working in the open
REWORK is led by Simon Cetin.
This website is developed through a supervised workflow that connects research, writing, design, development, testing and documentation. Changes are reviewed and versioned, and publication requires explicit approval from the responsible owner.
Reviewed changes.
Accountable publication.
A specific working problem
REWORK begins with a real task, a responsible owner and evidence that the problem is worth solving.
Explore it with Retoba