AI Implementation in 14 Days: What the Agent Actually Does

Monday, 9:20 a.m. The owner of a 60-person distribution company logs into a new system and finds a conversation instead of a settings screen. The first message from the agent reads: “Tell me which departments you have. Or just send me your staff list as a spreadsheet and I’ll sort it out.” Forty minutes later there’s a draft org chart on the screen: five departments, their heads, who reports to whom. One button at the bottom: Apply.

That’s what AI implementation in 14 days looks like at PRANUS. No six-month statement of work, no consultants on site. An AI agent does the configuration, and people make the decisions. Here is the two weeks, day by day: what the agent does, what you do, and what’s running at the end.

What “14 days” covers, and what it doesn’t

Let’s set expectations first. In 14 days you get a first working setup — the part of the business you run every single day:

• an org chart with departments, managers and reporting lines;
• permissions that follow from each person’s place in that chart;
• employees added and invited;
• sales pipelines in the CRM;
• tasks, lists and hand-off routes between departments;
• every customer conversation channel in one inbox;
• a knowledge base, sales scripts and answers to common questions, built from your own documents and chat history.

Complex manufacturing with bills of materials and costing, or a full replacement of your accounting system, is a separate phase with its own timeline. The ERP onboarding agent works the same way — by conversation — but we won’t promise you a factory in two weeks.

Who’s involved

The lead implementation agent runs the conversation, asks questions and assembles the configuration. Each service has its own specialist agent — org chart, CRM, tasks, documents, channels, knowledge base — and the lead agent hands work to them.

You, or a person you appoint, answer questions, upload files and say yes. That’s the core rule: the agent changes nothing without your approval.

Your employees each answer their own part. Your accountant sees a banner — “Implementation: your part” — with questions about accounting only. Your head of sales gets questions about the pipeline. Nobody has to read everything.

Every change follows the same five steps

1. Plan. The agent shows a card: what will change, before → after.
2. Your yes. One click, or a correction in plain words: “Logistics reports to the commercial director, not the CEO.”
3. Apply. The agent makes the change.
4. Verify. The agent checks the result against the plan.
5. Undo. Don’t like it? Roll it back in one action.

This keeps implementation from turning into a black box. At any moment you can see what was done, who approved it and how to reverse it.

Day by day

Days	What the agent does	What you do	What you get
1	Gets to know the company, asking one to three questions at a time with options and its own recommendation. Accepts any file: spreadsheets, policies, contracts, meeting recordings, voice notes	Answer, upload documents	Implementation dossier
1–2	Keeps the dossier: facts with their source (“who said it, which file”), open questions with an owner and a deadline, decisions made	Review and correct	Readiness map
2–3	Builds the org chart: departments and sub-departments with descriptions, managers, reporting lines, roles	Approve the plan	Live org chart in TEAMUS
3	Sets permissions: which services each department uses, at what level	Confirm	Access across every service
3	Adds employees: a “who goes where” list, invitations go out after your yes	Confirm the list	People in the system
3–5	Runs a distributed survey: questions assigned to specific people (“12 questions, about 15 minutes”), with reminders	Employees answer	Answers in the dossier
4–6	Configures services: pipelines and stages in CRMUS, task lists and routes, contract templates from Word, legal entities and accounts	Approve plans	Working pipelines and processes
4–6	Connects services: if a change affects another service, it flags the impact with ready-made next steps	Resolve loose ends	No broken links
6–8	Connects channels in CONNECTUS in a strict order: departments → pipeline → channel → routing → bot → a test message that must arrive as a lead	Enter login codes in a secure form	All conversations in one inbox
7–9	Imports conversation history	Say yes	Archive of customer conversations
8–10	Analyzes the history: strips personal data, filters out spam, extracts facts. Drafts documents for sales and support, plus a list of questions for the owner	Answer questions, approve drafts	The core of your knowledge base
8–10	Loads documents into MEMUS, connects the knowledge base to agents, tests it with control questions	Approve	A knowledge base that answers with a source
9–11	Prepares scripts: sales flow and objection handling from chats and calls	Approve the script	Scripts per pipeline stage
10–12	Turns on AI support and sales agents in “suggest, operator confirms” mode	Monitor and confirm	AI on the front line
11–14	Opens vacancies on empty seats in the org chart; the AI recruiter starts sourcing	Approve the salary range, run interviews	Hiring under way
14	Wraps up: readiness by module, a list of loose ends with owners	Sign off	A system in daily use

Treat the day ranges as a guide. A company with one pipeline and two channels moves faster; a group with five legal entities moves slower. The order of steps stays the same.

Where the agent asks instead of guessing

The implementation agent doesn’t pretend to read minds. It follows a simple rule: if the facts are missing, ask. If people say different things, ask.

Say your staff list shows Logistics reporting to the commercial director, but in a chat the logistics lead writes “I’ll check with the CEO.” The agent won’t silently pick one. It asks:

Who does Logistics report to? Options: the commercial director (per the staff list) or the CEO (per the chat history). I’d recommend the commercial director, so logistics tasks show up in his reports.

Unanswered questions don’t get lost. Each one lives in the dossier with an owner and a deadline. If nobody can answer, it escalates to the manager.

Where the scripts and knowledge come from

The most time-consuming part of a traditional rollout is capturing what lives in people’s heads: how to answer the delivery question, what to say when a prospect compares you to a competitor, what never to promise. Teams usually push it to “later,” and later never comes.

The agent builds it from what you already have. Your chat history with customers is a sales playbook — it just hasn’t been sorted. The knowledge factory reads through those conversations and drafts documents: services and pricing, tone of voice, sales flow, objections, FAQ, what not to promise, documents and processes. Support gets its own set.

We go deeper in Scripts and a knowledge base from your chat history.

What we need from you to hit 14 days

Research on implementations keeps landing on the same point: timelines slip because of people, not configuration. Panorama Consulting’s 2026 ERP Report names organizational issues — approvals, resistance, process redesign — as the top cause of schedule overruns. The agent takes over the busywork, but it can’t decide for you who owns what.

So before kickoff we agree on three things:

• One owner on your side who can say yes, or quickly ask someone who can.
• Two to four hours a week of a manager’s time for the agent’s questions — usually 15–20 minutes a day.
• Source material: staff list, policies, contract templates, access to work messaging accounts.

Why rollouts usually take months is a topic of its own — see Why ERP implementation takes months.

How this differs from “setup with an AI helper”

Plenty of systems now ship an AI assistant that tells you how to add a field or where a report lives. The human still designs the setup; the AI helps inside it.

PRANUS flips the roles. The agent builds the configuration — org chart, permissions, pipelines, channels, knowledge base — and a person reviews and approves. That works because every service is one system: a person is added once and appears everywhere, and permissions flow from the org chart.

What’s running after 14 days

• Every employee logs in and sees their own tasks, conversations and deals.
• Permissions are set by department, not person by person.
• Leads from every channel land in one pipeline.
• A new hire asks the knowledge base and gets an answer with a link to the source.
• AI agents answer customers by your rules and hand off to a person when they don’t know.
• Leadership has a map: what’s done, what’s left, and who owns it.

The bottom line

Fourteen-day implementation isn’t about rushing. It’s about the agent taking on the configuration work so people only handle what nobody else can decide for them. With an annual contract, implementation is free — details on the 14-day implementation page.

Want to see how the agent would talk to your company? Request a demo and we’ll walk through day one using your own setup.