Services
Explore

Free AI assessmentAnswer 15 questions about one process and see where AI could help.
Start an assessmentWhat should a 30-day AI plan for a small business include?
Week-by-week tasks, estimated time for each stage and the outcomes to check before moving on. Data and review rules come before any real work goes into a tool, and the month ends with a decision to continue, change or stop.
How To
How To
A week-by-week plan for a small business to test an AI tool on real work, with data and review rules agreed first and a decision to continue, change or stop.
A paid ChatGPT or Claude plan, chosen after you have checked the provider's current prices and data terms
About 90 minutes for the week 1 task audit
Roughly 2 to 3 hours a week for weeks 2 to 4
List the tasks that take the most time and note which involve writing, research or data. Check the data terms of two or three candidate tools and create accounts. Then write down which tool is approved, what information must not go into it and who checks the output before it is used.
Use the approved tool on your top task five times with real work that your rules allow. Check each output before it is used, and note how long the whole task took, including checking and corrections.
Turn the prompt that worked best into a reusable template, add standing instructions in your tool, and share the template, the data rules and the review step with anyone else who will use it.
Compare time and quality with your week 1 notes, write down what worked, and decide whether to continue, change the task or tool, or stop.
How the plan works
Intro
This plan is for a business owner with a full diary who wants to find out whether AI helps with a specific task in their business. It sets out what to do each week, roughly how long it takes and what to check before moving on.
Week 1 starts with a task audit on paper: you list what you and your team spend time on before choosing a tool. Later in the week you check the data terms of two or three tools, create accounts and agree three rules: which tool is approved, what information must not go into it, and who reviews the output before it is used.
Week 2 is the first trial on real work, within those rules. Week 3 turns what worked into a repeatable template and shares it with anyone else involved. Week 4 compares the results with your week 1 notes and ends with a decision: continue, change the task or tool, or stop.
Each week lists daily tasks, an estimated time, the outcomes to have by the end of it and a note on a common mistake at that stage. By day 30 you should have a written record of what the tool did well, where it needed correcting and whether it is worth continuing.
Weeks
Week1
GoalFind out where your time goes and agree how AI may be used before any real work goes into a tool. If you choose a task because it looks interesting, the trial is hard to measure later.
Days- Days 1 to 2 List your tasks Spend about 90 minutes writing down what you or your team do in a typical week. For each task, note how long it takes, how often it happens and whether it involves writing, research or data, which are the kinds of work general AI tools are most often used for. List everything before you filter. 90 min
- Days 3 to 4 Score and rank your tasks Score each task on two measures: time cost (hours per week in total) and suitability (is it repetitive, does it involve writing or research, and could you check the output easily?). Your top five become the list for the month. The highest-scoring task is the one you test first. 30 min
- Days 5 to 6 Check tools and create accounts Look for two or three tools that could handle your top task. Before you enter any business information, read each provider's current terms on how it uses and keeps what you enter, and check whether its business plan has different terms from the personal one. Create accounts for the tools you want to try. 45 min
- Day 7 Agree the rules and write your day 30 question Write three sentences: which tool is approved, what information must not go into it (for example client personal data or confidential documents), and who checks AI output before it is sent or used. Then write one sentence on what you want to find out: 'By day 30, I want to know whether [tool] reduces the time spent on [task], with output that is [usable after a light edit / usable as it stands / not yet usable].' Share both with anyone involved. 30 min
- A ranked list of your five most time-consuming tasks
- Accounts for two or three tools whose data terms you have checked
- Written rules on the approved tool, excluded data and output review
- A written day 30 question for your top task
Watch OutIf you pick a tool before you pick the task, you end up testing what the tool does well, which may not be what your business needs. Finish the audit first.
Week2
GoalUse the approved tool on real work, within your week 1 rules, and record what happens. This week is about regular use and clear notes. Polished prompts and workflows come later.
Days- Days 8 to 9 Use AI on your top task five times Each time your top task comes up this week, do it with the approved tool. Write a reasonable brief and see what comes back. Check each output before it is used, and note how long the whole task took, including checking and corrections. Your first prompts will be rough, and five imperfect attempts will teach you more than one carefully planned one. As you work
- Days 10 to 11 Try email drafting as a second use Use the approved tool to draft your next ten routine emails, keeping out any information your rules exclude. The question is whether editing a draft is quicker than writing from scratch. Note the answer for each email, because it can vary by type. As you work
- Days 12 to 13 Try one less obvious use Pick a task you were unsure about and try it once. For example: summarising a long document you received, drafting a proposal outline, writing a standard reply to a common client question, or preparing background notes before a meeting. Choose something low-risk, where a poor result costs little. 30 min experiment
- Day 14 Review week 2 Write three to five sentences: which tasks produced useful output, which did not, and what made the difference. Note the two prompts that worked best, because you will turn them into templates next week. If nothing was useful, write down why before you continue. 20 min
- Five AI-assisted attempts at your top task, each checked before use
- Notes on which prompts worked and which did not
- The time each attempt took, including checking
- Two prompts worth keeping
Watch OutOutput quality depends heavily on how the task is briefed. If the first result is poor, try a more specific brief before you decide the task is unsuitable. If several specific briefs still fail, that is a useful result too.
Week3
GoalTurn what worked into something you or your team can repeat. By the end of the week you should have at least one tested prompt template, standing instructions in your main tool and, if you have a team, a short note on how to use both.
Days- Days 15 to 16 Build your first prompt template Take the prompt that worked best in week 2 and write it out in full, with [placeholder] sections for the parts that change each time. Save it where you will find it, such as a pinned note or a shared document. Test it three times on real tasks and adjust the wording until the output is consistent enough to check quickly. 45 min
- Days 17 to 18 Set up standing instructions Most assistant tools let you save instructions that apply to each conversation, for example custom instructions in ChatGPT or a Project in Claude. Write a short description of your business, your role and how you write, so you do not have to repeat the same background each time. Apply your data rules to these instructions as well. 30 min
- Days 19 to 20 Share with your team (or add a second use if you work alone) If you have a team, choose one tool and one use, write a paragraph on how to use it, and attach the week 1 rules. For example: 'This is how I have been drafting client emails with [tool]. These are the rules on what can go in and who checks the result. Try it this week and tell me if it helps.' If you work alone, add a second use from your list, or test a connection between the tool and something you already use, such as your email. 30 min
- Day 21 Check the rules against what happened Read your week 1 rules again alongside your notes from weeks 2 and 3. Did anything come up that the rules did not cover, such as a new type of document or a question from a colleague? Update the rules and share the new version. 20 min
- At least one prompt template, written and tested
- Standing instructions set up in your main tool
- Team guidance shared with the rules attached (or a second use running if you work alone)
- Data and review rules updated after two weeks of use
Watch OutHold off on automation for now. Connecting AI to other tools is a step for after week 4, once a manual workflow has given consistent results that you have checked.
Week4
GoalCompare the results with your week 1 notes, write down what works and decide what happens next. Stopping is a valid outcome if the trial did not save time or the output needed too much correction.
Days- Days 22 to 24 Measure time against your week 1 notes Go back to your week 1 list. For each task you used AI on, compare how long it takes now with how long it took before, counting the time spent briefing, checking and correcting. For example, if a task took 45 minutes before and now takes 20 minutes including checks, that is 25 minutes saved each time. Multiply by how often the task happens to get a weekly figure. 45 min
- Days 25 to 26 Quality and reliability check For each use, are you happy with the output after review? How much editing does it usually need? Did any errors get past the review, or nearly reach a client? Is everyone checking in the same way? Aim for a review standard you can repeat and are comfortable with. Some editing is normal. 30 min
- Days 27 to 28 Write down what works Write a short note, half a page is enough, covering the tools you use, the two or three uses that worked best, your prompt templates and your data and review rules. You can give it to new team members, show it to clients who ask, or share it with an adviser. 45 min
- Days 29 to 30 Decide on month 2 Decide which uses to continue, which to change and which to stop. Check which tasks from your original list you have not tried. If the trial saved little or no time, decide whether to try a different task or tool, use ordinary automation, or keep the current process. Write up to three specific goals for month 2, or a note on why you are stopping. 30 min
- A measured time comparison for each task you tried
- A quality and reliability assessment of each use
- A short written note on tools, templates and rules
- A decision to continue, change or stop, with reasons
Watch OutIf the measurement shows little or no time saved, look at why. Was the task suitable? Was the brief specific enough? Did checking take longer than expected? Sometimes there is a fix. Sometimes the task is better done another way, and that is worth knowing too.
After day 30
What the end of month 1 can look like
If the trial went well, by day 30 you might have two or three tasks that take less time, one or two prompt templates you use each week, a short note you could hand to a new colleague, and a list of tasks from your audit still to try. If it did not, you have measured results showing where AI did not help, which is useful the next time a tool or vendor makes a claim.
In month 2 you can build on what worked, test whether one manual workflow is ready to connect to another tool, and look at sector-specific tools if they exist for your industry. Keep the data rules and the review step in place as you add uses.
The two or three uses that worked best in month 1. Improve the prompt templates and tighten the review step.
Take a manual AI workflow that has given consistent results and connect it to another tool through an automation service such as Zapier or Make, or a built-in integration. Start with the task that happens most often.
Pick a task from your original list that you did not reach in month 1 and apply the week 2 approach: try it five times, with checks, before building a process around it.
Common questions
FAQs
How much time do I need to commit to this plan?
Plan for roughly two to three hours a week. Week 1 needs a block of about 90 minutes for the task audit, plus time to check tool terms and agree the rules. In weeks 2 and 3, much of the work happens while you do tasks you would do anyway, so the extra time goes mainly on setting things up and keeping notes. Whether that time pays back is what week 4 measures. If you cannot find two hours a week at the moment, it may be better to start later.
What if I've already tried ChatGPT and it didn't help?
Poor results often come down to one of three things. The task may not suit AI: these tools are weaker on work that needs current information, specialist judgement or knowledge of a relationship. The brief may have been too vague, because a request such as 'write me a blog post about X' produces generic output. Or the output may not have been reviewed and refined, and drafts usually need editing. The week 1 audit is there to find tasks where AI is more likely to help. It is also possible that AI does not help with your tasks, and the plan is a way to find that out.
I'm a team of one. Does this still apply?
Yes. If you work alone, any time saved comes straight back to you, and decisions are quicker because nobody else needs to agree them. In week 3, skip the team-sharing step and add a second use or test a connection to another tool. Keep the week 1 data and review rules, because they matter just as much when you are the only person checking.
Which tools should I use in week 2?
Use the tool you approved in week 1. A paid ChatGPT or Claude plan covers most writing, research, summarising and drafting tasks. Free versions usually have tighter usage limits, which can make a short trial harder to judge. Check current prices and data terms on the provider's website, and note that consumer plans, business plans and the API can have different terms. Pick one tool for week 2 and use it consistently. Compare tools later, once you have results from one.
What counts as a good result by day 30?
A good result is a clear answer to the question you wrote on day 7. Useful checks: Did your top task take less time, counting briefing and checking? Do you have at least one prompt template you use regularly? Does the output meet your review standard without heavy editing? Do you have written rules on tools, data and review? If the answers are mostly no, look at whether the task suited AI and whether the brief was specific enough. If they are still no after that, stopping or trying a different task is a reasonable decision.
What happens in month 2?
Month 2 is for building on the two or three uses that worked best in month 1, adding one or two new uses from your original list, and checking whether any workflow is ready to connect to another tool through Zapier, Make or a similar service. It is also a good time to look at sector-specific AI tools if they exist for your industry, and to compare them with a general assistant on the same task.
Talk to us
Our AI workshops cover the early stages of this plan with your team
An AI workshop can cover the task audit, tool selection, prompt building and planning how your team will use AI, using your own business as the example. The trial on real work and the measurement then run over the following weeks.