Before You Start
Too many businesses jump into AI without a plan. They sign up for tools, experiment randomly, get frustrated, and give up. This checklist gives you a structured path from "we should probably look at AI" to "AI is saving us real time and money."
Phase 1: Planning (Steps 1 to 5)
Step 1: Define the Problem
Write down the specific business problem you want AI to solve. Not "use AI to improve our business" but "reduce customer response time from 4 hours to under 30 minutes" or "cut our content production time by 50%."
If you cannot define a specific problem, you are not ready for AI. Go back and map your repetitive tasks first.
Step 2: Estimate the Value
Put a rough number on it. If the problem costs your business 10 hours per week at £15/hour, that is £7,800 per year. An AI solution that costs £100/month (£1,200/year) would deliver a 6.5x return. This gives you a budget range and a way to measure success.
Step 3: Audit Your Data
What data do you have that is relevant to the problem? Where is it stored? Is it clean and structured, or messy and scattered?
AI needs data to work. If your data is in paper files, scattered across personal email accounts, or inconsistently formatted, you may need to clean it up first.
Step 4: Check Your Team's Readiness
Talk to the people who will use the AI solution. Are they open to it? Worried about it? Excited? Understanding their mindset helps you plan training and change management.
Step 5: Set a Timeline
Give yourself a realistic deadline. For simple AI tool adoption (like ChatGPT for email drafting), 2 weeks is enough. For a custom AI chatbot or automation project, plan for 4 to 8 weeks.
Phase 2: Selection (Steps 6 to 10)
Step 6: Research Solutions
Based on your defined problem, research available solutions. Check:
- Off-the-shelf AI tools that address your specific need
- AI features built into software you already use
- Custom solutions (if off-the-shelf does not fit)
Start with the AI tools guide for options.
Step 7: Compare Pricing
Get clear pricing from at least two options. Check for:
- Monthly vs annual pricing
- Per-user vs flat-rate pricing
- Usage limits on different plans
- Hidden costs (setup fees, integration fees, training costs)
Step 8: Check GDPR Compliance
If the AI tool will process personal data, verify:
- The provider offers a Data Processing Agreement (DPA)
- Data is stored in the UK or EU, or adequate safeguards are in place
- Your privacy policy covers AI tool usage
See the GDPR and AI guide for details.
Step 9: Request a Trial or Demo
Never commit to a paid tool without testing it first. Most AI tools offer free trials or free tiers. Use them with real (but non-sensitive) business data to evaluate quality.
Step 10: Get Buy-in
Share your plan with anyone who will be affected. Explain the problem, the proposed solution, the expected benefits, and the timeline. Address concerns honestly. If key people are not on board, implementation will struggle.
Phase 3: Implementation (Steps 11 to 15)
Step 11: Start Small
Do not roll out to the entire business at once. Pick one team, one department, or one process. Prove the value there first.
Step 12: Configure and Customise
Set up the tool with your specific data, branding, and preferences. For a chatbot, load your FAQ and knowledge base. For an automation tool, configure your specific workflows. For an AI assistant, set up custom instructions with your brand voice.
Step 13: Train Your Team
Run a focused training session. Cover:
- What the tool does and does not do
- How to use it for their specific tasks
- Common mistakes to avoid
- Who to ask for help
- Your AI usage policy
Step 14: Run a Pilot Period
Let the tool run for 2 to 4 weeks with close monitoring. Collect feedback from users daily in the first week, then weekly.
Step 15: Fix Issues Quickly
Problems will surface during the pilot. A chatbot gives a wrong answer. An automation triggers incorrectly. A team member struggles with the tool. Address these immediately. Quick fixes build trust; slow responses erode it.
Phase 4: Optimisation (Steps 16 to 20)
Step 16: Measure Against Your Goal
Go back to Step 1. Has the problem improved? Compare your before and after metrics:
- Time saved per week/month
- Cost savings
- Quality improvements
- Customer satisfaction changes
- Error rate changes
Step 17: Gather Feedback
Ask your team: What is working? What is frustrating? What could be better? What else could we automate?
Step 18: Optimise
Based on metrics and feedback, refine the implementation. This might mean:
- Updating the knowledge base for a chatbot
- Adjusting automation triggers
- Adding new use cases
- Upgrading to a higher tier for better features
Step 19: Document Everything
Write down how the solution works, how to maintain it, and how to troubleshoot common issues. This ensures the solution survives even if the person who set it up leaves.
Step 20: Plan Your Next AI Project
Once the first project is delivering consistent value, repeat the process for the next highest-priority problem on your list. Each project gets easier as your team builds confidence and expertise.
Common Reasons AI Projects Fail
- No clear problem to solve. Vague goals produce vague results.
- Poor data quality. AI cannot work with messy, incomplete data.
- No team buy-in. If people do not use the tool, it does not matter how good it is.
- Trying to do too much at once. Start small, prove value, then expand.
- No measurement. If you do not measure results, you cannot prove value or identify problems.
- Giving up too soon. Most AI tools need 2 to 4 weeks of tuning before they perform well. Do not abandon after 3 days.
Frequently Asked Questions
How long does a typical AI implementation take?
For off-the-shelf tools (like ChatGPT or a basic chatbot): 1 to 2 weeks. For custom automation: 4 to 8 weeks. For bespoke AI applications: 8 to 16 weeks.
Do I need a consultant for implementation?
For simple tool adoption, no. For custom automation, integration with existing systems, or anything involving sensitive data, a consultant will save you time and reduce risk. Book a free discovery call if you want to discuss your situation.
What is the minimum budget for an AI project?
You can start for free with ChatGPT Free or Claude Free. A meaningful pilot with paid tools typically costs £50 to £200/month. Custom implementation projects start from around £2,500.
Should I build or buy?
Buy first. Off-the-shelf tools cover most small business needs at a fraction of the cost of custom development. Only build custom when you have a specific requirement that no existing tool addresses.