Most AI assistants today work as a separate window that employees type questions into. They can draft an email, summarise a document or suggest a next step, but the output stays in that window until a person copies it somewhere else. They have no access to the company's email, CRM or ERP, they don't know the company's internal rules, and they can't take an action on anyone's behalf. As a result, an employee still has to move the information between email, the CRM, the ERP, spreadsheets and internal tools by hand. At Cadr, we think this is the main reason AI has done less to change day-to-day operations than most companies expected.
Cadr coworkers are built to work inside the company instead. Each coworker is assigned to a specific employee and is set up with that person's role, the tools they use, the processes they are responsible for and the permissions they hold. It can read and write in the same systems that employee uses, and it follows the same approval rules that apply to them. In practice it behaves like a new team member who has been given system access, a copy of the company handbook and a clear list of what they can sign off on alone. Four features make this work in a real business, and each one addresses a specific reason companies have been reluctant to let AI take actions.
Permissions set by role and risk
A coworker that can take any action without limits would be a risk to the business. Before it handles real work, it needs a defined list of what it can do independently, what requires approval and what it can never do without explicit permission. Cadr lets companies set these rules for each role, so the level of independence matches the risk of each action. Routine, low-risk tasks can run automatically, while anything involving customers, pricing or commitments goes to a person for review.
A sales team shows how this works. An outbound sales coworker can identify prospective clients, research them and prepare personalised outreach, but it can't send the first email until a manager has approved both the account and the message. A sales operations coworker can draft a quotation using approved pricing rules, and the salesperson reviews it before it is sent. A coordination coworker can update the CRM automatically from an email thread, but if a customer asks to change a delivery date, it sends that request to the responsible manager instead of updating it alone. Companies have several levels of authority, and the cost of an error differs from one action to the next, so the coworker's permissions are set to reflect both.
Trained on company policy
Every company has its own operating rules, and most of them are internal documents that a general AI model has never seen. These include pricing policies, discount thresholds, approval chains, customer communication standards, escalation procedures, compliance requirements, training manuals and informal conventions that teams follow without writing down. An assistant that doesn't know these rules produces work that someone has to check and correct. Each correction costs time, and repeated errors make teams less willing to rely on automation at all.
Cadr coworkers are trained on a company's own policies and operating manuals, so the work they produce follows the company's procedures. If a customer hasn't replied to a quotation, the coworker sends follow-ups on the schedule the company has approved. If a discount above a set threshold needs CFO approval, the coworker requests that approval before the quotation is sent to the customer. If a certain type of customer requires additional compliance checks, the coworker runs those checks on every deal, with no exceptions for deadlines or workload.
Completing tasks in the company's systems
Most AI tools today report what needs to be done and leave the work to the employee. A tool might point out that a CRM record is out of date, but someone still has to open the CRM and make the change. That means the AI adds a step to the employee's day without removing one. Cadr coworkers are designed to make the change themselves, within the permissions they've been given.
For example, when a customer sends an enquiry by email, a Cadr coworker can create a CRM entry with the opportunity value, product requirements and expected closing date filled in from the message. It can then prepare a PDF quotation from the company's approved templates, create a follow-up task, update the matching ERP record and send the salesperson a short list of the points that need their decision. The salesperson keeps control of pricing, terms and the final send. The coworker handles the data entry and the transfers between systems that would otherwise take up part of their day.
The time saved increases with the number of systems a role uses. A salesperson might work across email, the CRM, WhatsApp, spreadsheets and an ERP system in a single day. A procurement manager handles supplier emails, purchase orders and inventory software, and often enters the same figures into each one separately. A CEO relies on reports that several staff members put together by hand from different sources. Cadr coworkers work across all of these systems while applying each employee's permissions and the company's policies at every step.
A full record of every action
If AI is going to handle real business processes, managers need a clear record of what it has done. Cadr logs every action a coworker takes, keeps permission settings under the company's control and lets managers review each coworker's communication with customers and internal teams. For any task, a manager can see what was done, the reason for it, and whether each step was automatic or approved by a person. This gives leadership the same oversight of a coworker that they would expect over any employee.
Because Cadr connects email, CRM activity and operational systems, it can also report on patterns that are hard to see when the data sits in separate tools. It can show at which stage customers leave the sales funnel, which quotations take longest to be approved or answered, where follow-ups were not sent and which accounts have reduced their activity. It can also highlight areas where ERP data shows unusual costs or waste. Producing this kind of analysis by hand usually takes an analyst several days of gathering data from different departments.
How Cadr changes the way companies use AI
Cadr places AI inside a company's existing systems and processes, so it is no longer a separate tool that employees have to consult. Each coworker is set up for a specific person's job, works in the same software they use, follows the company's written policies and operates within the approval structure already in place. It completes tasks from start to finish, and decisions about pricing, commitments and customer relationships stay with the people responsible for them. This is the product we are building at Cadr.