If you've spent any time reading about AI in the last year, you've probably noticed that the coverage tends to swing between two extremes. Either AI is going to solve every problem your business has ever had, or it's a dangerous hallucination machine that will embarrass you in front of your best clients.

Neither is accurate. Here's a more useful version: five things AI genuinely does well right now, and five things that still require a human being.

Five things AI is genuinely good at

1. First drafts of almost anything written

Emails, proposals, job ads, social posts, website copy, staff policies, client updates, tender responses. If it starts with a blank page and ends with words, AI can give you a solid 80% draft in seconds. The remaining 20% — your voice, your specific context, your judgement call — still needs you. But starting from a draft rather than nothing is a significant time saving.

2. Summarising long documents

Paste in a 30-page contract, a lengthy supplier proposal, or a stack of meeting notes, and AI will give you a clear summary in under a minute. It's genuinely useful for anyone who spends time reading documents they'd rather not read in full. The caveat: always check the summary against the original for anything high-stakes. AI can miss nuance.

3. Cleaning and organising data

Messy spreadsheets, inconsistent client lists, data exported from one system that doesn't quite match the format of another — AI handles these well. Claude can reformat, deduplicate, sort, and standardise data with a well-constructed prompt. Tools like Microsoft Copilot can do this directly inside Excel.

4. Generating options and alternatives

Stuck on what to call a new service? Need five different ways to phrase a difficult email? Trying to come up with ideas for a promotion? AI is an exceptionally fast brainstorming partner. It won't always give you the perfect answer, but it will give you ten options to react to instead of a blank page to fill. That's often more valuable.

5. Answering questions about documents you've given it

You can paste in a document — a contract, a policy, a report — and ask Claude specific questions about it. "What are the payment terms in this agreement?" or "Does this document mention anything about liability?" It's a fast way to find information in long documents without reading every word yourself.

The common thread across all five: AI works best when you give it something to work with. The more context you provide, the more useful the output. Think of it as a fast, capable colleague who needs to be briefed — not a mind reader.

Five things AI still can't do

1. Know your business without being told

AI tools have no memory of who you are or what your business does unless you tell them in every conversation. Every session starts from scratch. This means the quality of what you get back depends heavily on the quality of the context you provide. Give it nothing, get generic output. Give it detail, get something genuinely useful.

2. Guarantee accuracy on facts

AI hallucinates. That's the technical term for when it produces confidently-worded information that is simply incorrect. It doesn't know it's wrong. This is a genuine limitation for tasks that require factual precision — tax figures, legal clauses, technical specifications. Use AI to draft and organise; verify the facts yourself or with the relevant expert.

3. Replace relationship and judgement

What most small businesses actually sell is trust, expertise, and relationship — not just the deliverable. An AI can write a proposal, but it can't read the room in a discovery meeting, recognise that a client is nervous, or make the call to walk away from a job that isn't right. That judgement is yours, and it's worth more than people often realise.

4. Operate independently without your input

The version of AI where software runs your business autonomously in the background is not the version that exists today. Current AI tools respond to prompts. They don't initiate, plan, or manage themselves. You are still the operator. You ask, it responds. You review, you send. It's a tool, not an employee.

5. Understand your industry context automatically

A general AI tool doesn't know the norms of your industry, the specific language your clients use, or the particular way your trade or profession operates. It will write you a proposal in the general style of the industry, but it won't capture the specific things that make your business distinctive unless you explain them. This is exactly the gap that tools built specifically for particular industries are designed to close — and the kind of work Wolfari specialises in.

The honest summary

AI right now is a genuinely useful productivity tool for small businesses, particularly for language-based, repetitive, and administrative tasks. It is not a replacement for expertise, relationships, or human judgement. The businesses getting the most value from it are the ones treating it like a fast, tireless assistant — not a decision-maker.

The best way to develop an accurate picture of what it can do for your specific business is to try it. Start with something low-stakes — a draft email, a summary of meeting notes, a few options for a social post. See what comes back. Your instincts about what's useful and what's not will sharpen quickly.