A Roadmap to AI-First Without Pushback From Your Team

A CEO of a 45-branch restaurant chain in Saigon messaged MONA last week: “I bought Claude Team. But my staff are pushing back. Can you suggest a week-by-week roadmap?” He is not the only case. We work with nearly 30 mid-market Vietnamese businesses we know well. More than half of the CEOs get stuck in the first 30 days. After buying the tool, staff either avoid it or get irritated. They copy ChatGPT content straight into Word and then say “AI is useless”. In this article MONA shares a 30-day CEO AI roadmap broken down by week. You can start running it next week and measure by how many tasks staff can actually hand to AI. See more how employees react when you roll out AI.

Why 90% of staff push back when using AI

The story of the restaurant chain CEO above isn't unusual. We recognize a familiar pattern: the CEO buys AI and hands it straight down to staff without using it first. Staff get the order to “apply AI” but have no prompt templates. There is no standard brand voice and nobody to guide them. After 2 weeks of fumbling, everyone goes back to the old way.

The second reason. You often expect AI to replace 50% of staff in the first month. The truth: AI replaces 60-90% of repetitive work over 6-12 months, with a layer of people in between to supervise. The State of AI 2025 report by McKinsey shows that 78% of businesses worldwide have tried at least one AI use case, but only 22% moved to running it for real. The rest are stuck at “tried it and left it”.

The third reason, and the heaviest. The CEO brags about AI on LinkedIn, then doesn't sit down to review staff output in the first 2 weeks. Subordinates guess, produce poor quality content, and then review it by hand 2-3 times. The total cost is higher than without AI. This is the pattern we analyzed in the article CEO sài AI across 4 specific wrong roles.

The fourth reason. You have no integration layer. AI stands alone (ChatGPT open in the browser, Claude open in an app), so staff copy back and forth, lose context and lose discipline. To really cut time, AI has to connect to the CMS, CRM, accounting and storage. This is the part where MONA writes custom automation software for each client. We don't use drag-and-drop tools like n8n or Zapier.

Week 1: the CEO uses AI every morning before meetings

In week 1 you do NOT touch staff. The only rule this week: use AI for 30 minutes every morning on 1 real task of the boss. No meetings about AI, no announcements, no bragging on LinkedIn.

Which tasks specifically? Pick 1 of the 3 below, and change no more than 2 times during the week.

Task one in week 1 of the CEO AI roadmap: summarize the revenue report

The boss pastes the Excel file or PDF report into Claude and gets 5 key points. Do it for 5 minutes every morning. After 1 week, you see for yourself which points are useful and which are empty. That is real experience, not reading hype.

Task two in the first 30 days of AI: prepare content before the board meeting

An hour before the meeting, the boss feeds the agenda and 5-10 related documents into ChatGPT or Claude and gets a 1-page summary and 3 suggested questions. Your meetings get up to 70% less tiring.

Task three for a Vietnamese CEO using personal AI: weekly pricing analysis

The boss feeds revenue data by SKU into Claude and asks “suggest 3 pricing options for the 2 SKUs that aren't selling”. AI returns 3 options with reasons. The boss picks 1 for staff to test.

By the end of week 1, you have 2 things: personal experience (the boss can use it now), and 3-5 prompt templates that match the boss's brand voice. These two things are your weapons for talking to staff in week 2 without getting into arguments.

Week 2: pick 1 department and 1 process to test AI

In week 2 you do NOT roll AI out to the whole company. The rule: 1 department, 1 process, 2 people. The goal is an experiment with measurable results in 7 days, not “applying AI across the whole business”.

A good process to test AI in week 2 must answer the 4 questions below, all with “yes”.

Question one: does this process run at least 5 times a week?. Run less often and you won't have enough data to measure improvement. High-frequency processes like replying to customer messages and drafting outreach emails work well. Screening CVs, writing product descriptions and summarizing sales calls are all fine too.

Question two: does the result have a clear format? Replying to messages from a template. A 5-line outreach email. CVs by 3 fields (experience, skills, education). Vague processes like “advising big customers” or “drafting strategy” should NOT be picked in week 2.

Question three: do 2 people in the department agree to be the “test group”? Do NOT force the whole department. Pick 2 people who are curious, learn fast and aren't afraid to try new things. We see that staff aged 25-30 usually get used to it faster than the 40+ group. Tell these 2 people clearly: “this is a 7-day experiment. It doesn't affect the old performance metrics, and the results are shared with the whole department”.

Question four: can you measure the result within 7 days? For example, replying to messages: measure the time from when the customer messages to when they get a reply. Outreach emails: measure the reply rate. CVs: measure how many pass screening. You need numbers before and after, not “gut feeling”.

Week 3: set specific metrics for AI, nothing vague

In week 3 you have real data from the 2 people in the test group. Now sit down with the department head and the 2 testers and set the metrics for the next 7 days. A metric must have a number, a timeframe and a person accountable.

In the 30-day CEO AI roadmap, we see a common mistake: the CEO writes a metric like “applying AI to raise marketing department productivity”. This metric is NOT measurable. Staff don't argue, but they don't do it either. After 2 months, the CEO says “AI isn't effective”.

A correct metric looks like this: “reply time to customer messages drops from 4 hours to 1 hour 35 minutes in 30 days, Ms. Trang in charge”. This is the metric the CEO of the 45-restaurant chain in Saigon set for the customer care department. In March the result was 1 hour 28 minutes, beating the target.

Or another form: “the number of CVs passing screening rises from 12 to 30 out of 200 CVs per hiring round, HR Lan in charge”. A COO of a 28-branch spa chain across HCMC, Hanoi and Da Nang used this metric. In February the result was 28 CVs passing, saving 18 HR hours per hiring round.

In week 3, you review the test group twice (midweek and at the end of the week), 30 minutes each time. We find the boss's 30 minutes of review is worth more than 5 training sessions. Review is when the boss spots where AI output is off the brand voice and fixes the prompt templates right away.

Week 4: expand AI to the next 2 departments

In week 4 you do 3 things in the following order.

Thing one: review the week 3 numbers with the board. Bring the metrics to a 30-minute board meeting. Show the good numbers and show the bad ones too. The goal: the board supports the expansion, not to show off. We see that a CEO who shows bad numbers early earns the board's trust for longer than one who only shows good numbers.

Thing two: pick the next 2 departments. Base it on the process the test group standardized in week 3. For example, if the test group was customer care, expand to sales (replying to outreach emails) and marketing (writing product descriptions). Avoid complex departments (finance, legal) for the second wave.

Thing three: write a 1-page internal handbook. The handbook includes: prompt templates, brand voice rules, the new metrics and the person in charge. Keep it in Notion or an internal Google Drive. Each department reads it for 15 minutes and starts running.

By the end of week 4, you have: 1 test department that has run AI for 30 days, 2 more departments expanded, and 1 short internal handbook. This is real adoption, not “signing up for Notion AI and leaving it there”.

5 mistakes to avoid when rolling out AI

We recognize 5 mistakes Vietnamese CEOs often make in the first 30 days. Avoiding them saves 6 months of trial and error.

Mistake one: buying many tools at once. You sign up for ChatGPT Team, Claude Pro, Manus, Notion AI and Perplexity Pro all in week 1 “just to try”. The result: 30 days pass and no tool is used deeply enough to build a standard template. The rule: 1 main tool for the whole 30 days, switch after the week 4 review. See the details in the article the AI First handbook for businesses.

Mistake two: making a whole department use AI in week 1. The boss sends an email saying “from this week the marketing department must apply AI”. The whole department panics, and either avoids it or copies AI output straight in. Without a test group there is no standard template. You have to start with 2 people.

Mistake three: vague metrics like “applying AI to raise productivity”. As said in week 3. A metric must have a number, a timeframe and a person accountable.

Mistake four of Vietnamese CEOs using AI: laying off staff in the first 30 days of AI

Laying off early is the most expensive mistake a Vietnamese CEO makes with AI in the first week. Last month, the CEO of a 12-store gold chain in HCMC planned to lay off 4 customer care staff after 2 weeks of trying AI. We had to advise him to stop. The reason: 2 weeks is not enough to prove AI can replace them. Laying off and rehiring costs 3 times the salary saved. Wait until after 90 days to decide on staff.

Mistake five when rolling out AI in a business: no integration layer with the existing software

The integration layer is what MONA always prioritizes in real business automation. You buy AI but let staff copy back and forth between the CRM, CMS and accounting. After 30 days, each employee loses 30-40 minutes a day just copying. This is the point where MONA usually steps in to deploy custom automation software, tying AI into a closed chain of modules.

Next steps and support tools after 30 days

After 30 days, you have 3 options for expanding further.

Option one: keep expanding with the tools you already have. You keep ChatGPT Team or Claude Team and write more handbooks for the next 3-5 departments. This suits businesses with fewer than 100 staff and processes that aren't too complex.

Option two: hire MONA to write custom automation software. This suits you if you have 200+ staff, 15+ branches, or need to tie AI into an existing CRM, ERP or LMS. We have done this for nearly 30 Vietnamese mid-market businesses. The article MONA automates 30 businesses shares the specific patterns.

Option three: combine existing tools with small modules MONA writes for you. You keep Claude Team for marketing, and MONA writes an automation module for accounting or customer care. This hybrid approach saves upfront cost and suits businesses with 50-200 staff.

You can pick any option you like. We suggest MONA only after you have run the 30-day test group and have specific numbers. Don't hire automation software before you know which processes can be automated.

The same pattern applies to 6 mid-market industries

The 30-day roadmap we shared above isn't just for restaurants. We have rolled it out for 6 mid-market industries with light adjustments for each one.

The AI roadmap for CEOs of spa, salon and nail chains

The test group is usually customer care or booking. The test process: replying to customers on Zalo and automatic appointment reminders. MONA did this for the 28-branch spa chain across HCMC, Hanoi and Da Nang with spa and salon management software AI integration.

Industry two: restaurant, café and F&B chains. The test group is usually the shift supervisor team in operations. The test process: summarizing end-of-shift revenue reports and suggesting next week's menu. Integrated restaurant management software through the Nhahang AI module.

Industry three: fashion and cosmetics retail chains. The test group is usually marketing and sales. The test process: writing product descriptions and replying to Facebook Page messages.

AI rollout for education centers and online academies

The test group is usually academic affairs and customer care. The test process: drafting follow-up emails to parents and summarizing the weekly student report. MONA has MONA EduCenter software with an AI assistant integrated.

Industry five: transport and importing goods from China. The test group is usually the dispatch team. The test process: automatically sorting orders and flagging delays. MONA has software for importing goods from China that is already running for real.

Industry six: real estate, boarding house rental and brokerage. The test group is usually sales and customer care. The test process: filtering Facebook Ads leads and replying to customers asking on Zalo. Integrated boarding house management software.

The first 30 days are only the start with AI

When you finish the 30 days, don't think “we've used AI”. This is only the start. After a good 30 days, we usually suggest the next 12-month stretch: months 2-3 expand to 4 more departments, months 4-6 write custom automation modules, months 7-12 re-measure all the company's metrics. See the details in the article the AI First handbook for businesses.

If you want us to work alongside you through the first 30 days, message MONA via Mona Software. We have a 30-day advisory package for CEOs. It includes 4 one-on-one sessions with one of the MONA founders who have built real AI software, plus a handbook tailored to your industry.

We are entering a phase of decisive automation together. MONA has run it for ourselves since April 20, and every day feels like a new company. So who among you will start your 30-day CEO AI roadmap next week?

Frequently asked questions about the 30-day CEO AI roadmap

Is the 30-day CEO AI roadmap different from the common 90-day roadmap?

The difference is that MONA packs the 30 days just for the CEO, not for the whole company. 30 days is just enough for the CEO to use AI each morning. Pick 1 department to test, measure with 3 metrics of money, hours and errors, and then expand. After 30 days the CEO has enough data to decide the 12-month roadmap.

Does a Vietnamese CEO need to know how to write prompts before starting the 30-day roadmap?

Not required. In week 1, the CEO only needs to open ChatGPT or Claude and type questions as if asking an assistant sitting next to them. After 3-5 days you will see for yourself which phrasing gives good results. MONA doesn't advise studying prompt engineering first, because learning by hand is remembered longer.

Which department should be picked to test AI in week 2?

We recommend a department that has data ready, a clear process, and a head who is open to it. Usually customer care or content marketing. Avoid departments in internal conflict or restructuring. For this step MONA often sits down for 60 minutes to help the CEO decide.

If the metrics don't look good after 30 days, continue or stop?

If after 30 days 1 of the 3 metrics of money, hours and errors hasn't improved, don't stop in a hurry. Go back over the 5 common mistakes we listed. In 80% of cases it is the wrong use case or measuring the wrong way, not a dumb AI.

How much do the first 30 days of a CEO using AI cost a mid-market company?

A ChatGPT Plus or Claude Pro subscription is about 20-25 USD per CEO. Adding 1 department to test an AI agent, the 30-day budget runs about 5-15 million VND for a company with 200-500 staff. MONA has seen most businesses save 5-10 times that amount in the very first month.