I tried 5 AI tools so you don't have to: an honest small business review
Not a sponsored roundup. Not a listicle written by someone who opened each tool for 10 minutes. These are the tools I actually used for real business tasks — here's what worked, what didn't, and what I'd start with today.
Every week there's a new article ranking "the best AI tools for small business." Most of them were written by someone who spent twenty minutes with each tool, copy-pasted the feature list from the pricing page, and added an affiliate link at the bottom.
This isn't that.
The five tools reviewed here were tested against real small business tasks over an extended period — not demos, not toy prompts, but actual work: writing customer emails, drafting proposals, building FAQ content for a chatbot, analysing a monthly P&L, producing a week of social media content, researching competitors, and generating job descriptions. The kind of work a real business owner does every week.
The format for each review is consistent: what the tool is actually for, what it does well on real tasks, where it falls short, what it costs, and who should use it. At the end there's a straight recommendation based on business type and budget — including the one tool to start with if you're new to all of this.
One disclosure: none of the tools reviewed here have paid for placement, offered discounts in exchange for coverage, or been given any editorial input. AInstein doesn't do sponsored reviews. The whole point of this publication is to be the resource you can trust when everyone else is trying to sell you something.
How I tested these tools
Each tool was used for a minimum of four weeks on real tasks. The tasks were consistent across tools where comparison was meaningful — the same brief given to multiple tools to produce a social caption, the same financial data given to multiple tools to produce a summary.
Where tools serve completely different purposes (a chatbot builder vs a writing assistant), the comparison is within category rather than head-to-head.
The business context used for testing: a fictional but realistic five-person service business — a marketing agency with a mix of retainer and project clients, a small customer list, an active social media presence, and the usual admin overhead of a growing small business. Results will vary for different business types — where relevant, those differences are noted.
Tool 1: Claude (Anthropic)
Category: General AI writing assistant and reasoning tool Tested for: Content creation, email drafting, financial analysis, proposal writing, SEO research
What it actually is
Claude is a conversational AI assistant built by Anthropic. In plain terms: you type something, it responds. The difference between Claude and a search engine is that Claude doesn't give you a list of links — it thinks through your question and gives you a complete answer, or produces the thing you asked it to produce.
It's best understood as a very capable writing partner and thinking tool. Not a search engine. Not a database. A collaborator.
What it does well
Writing quality is consistently the best of any tool tested. When given a detailed brief — including a voice document, audience description, and specific task — Claude produces first drafts that require less editing than any other tool. The output doesn't just follow instructions; it seems to understand why you're doing something and writes accordingly. Emails feel like they came from a person. Blog posts have an argument, not just information arranged in sections.
Complex reasoning tasks are handled well. Giving Claude a P&L report and asking it to explain what's happening in plain English — flagging concerns, identifying trends, suggesting questions to ask an accountant — produces output that's both accurate and genuinely readable. This is harder than it sounds. Most AI tools either oversimplify financial summaries or produce ones that are technically correct but practically useless.
Long-form consistency. For articles, proposals, and longer documents, Claude maintains a consistent voice and argument across the full length in a way that some other tools don't. It doesn't introduce new tones or contradict itself halfway through.
Following specific instructions. When you tell Claude not to use certain phrases, to keep something under a specific word count, or to write in a particular format, it follows those instructions reliably. This sounds basic but it's where many tools fail — they acknowledge the instruction and then ignore parts of it.
Where it falls short
No real-time web access on the standard model. Claude's knowledge has a cutoff date, which means it can't tell you what happened last week or give you current pricing from a competitor's website. For tasks that require current information — market research, competitor analysis, checking recent news — this is a real limitation. Claude does have web search capability in some configurations, but it's not the default experience.
Image generation is not available. If you need to create visual content alongside your written content, Claude can't do that. You need a separate tool (Canva AI, Adobe Firefly, or Midjourney) for images.
The free tier has daily limits that active users hit quickly. The limits reset, but if you're using Claude as a daily work tool and hitting restrictions mid-workflow, it's disruptive.
Pricing
Free tier: Claude Sonnet with daily usage limits — functional for occasional use. Claude Pro: $20/month — removes limits, adds access to Claude Opus (the most capable model), and unlocks Projects (saved workspaces that remember your context across sessions).
Verdict
Best for: Any business owner doing regular writing tasks — content, emails, proposals, documentation, financial summaries.
The Projects feature on the Pro plan is the most underrated feature in this entire review. You create a workspace, paste in your voice document and business context once, and every subsequent conversation in that Project already knows who you are and how you write. For content-heavy businesses, this alone justifies the $20/month.
Rating: 9/10 — the best writing tool tested. The web access limitation is the only meaningful gap.
Tool 2: ChatGPT (OpenAI)
Category: General AI assistant, research, and custom tool building Tested for: Research, writing, custom GPT building, image generation, web browsing tasks
What it actually is
ChatGPT is the tool that made AI accessible to mainstream audiences and still has the largest user base of any AI assistant. It's a conversational AI tool with a broader feature set than most competitors — including real-time web browsing, image generation via DALL-E, and the ability to build custom configurations (custom GPTs) that are pre-loaded with your instructions.
What it does well
Real-time web access is the meaningful differentiator from Claude on the standard free-to-paid tier. When you need current information — checking a competitor's pricing, finding recent statistics, researching what's happening in your industry this month — ChatGPT can browse the web and bring the results into its response. For research-heavy tasks, this is a genuine advantage.
Custom GPTs are powerful for business owners willing to invest the setup time. A custom GPT is essentially a pre-configured version of ChatGPT with your instructions, voice document, and business context baked in. You build it once and then anyone on your team (on a paid plan) can use it without having to know how to write prompts. A custom GPT called "write a customer email" that already knows your business and voice is more accessible to non-technical team members than explaining prompt engineering.
Image generation. DALL-E integration means ChatGPT can produce images for social media, blog post headers, and basic visual content. The quality is functional rather than exceptional for marketing purposes — for professional imagery, dedicated tools like Midjourney or Adobe Firefly produce better results — but for quick visual content without a separate tool, it's genuinely useful.
Breadth of use cases. ChatGPT has been trained and refined on a wider range of tasks than most competitors. Code, data analysis, spreadsheet formulas, research summaries, creative writing, voice interaction — the breadth is wider than any other tool in this review.
Where it falls short
Writing quality is slightly below Claude for nuanced tasks. When given the same detailed brief, ChatGPT's output tends to be competent rather than excellent. It follows instructions but doesn't always seem to understand the purpose behind them. Emails and articles produced by ChatGPT need slightly more editing to reach the same quality bar. This gap narrows significantly with very detailed prompts but it exists.
Inconsistent behaviour across sessions. ChatGPT occasionally ignores instructions it acknowledged, produces different quality output on similar tasks in different sessions, and seems to perform differently depending on server load or the time of day. This inconsistency is less pronounced on the paid plan but it's a real characteristic.
The free tier is increasingly limited. As ChatGPT has grown, the free tier has become less generous. Usage limits are tighter than they used to be, and some features (web browsing, DALL-E) are restricted or unavailable on the free plan.
Pricing
Free tier: GPT-4o with usage limits, some features restricted. ChatGPT Plus: $20/month — higher limits, consistent access to the latest model, web browsing, DALL-E, custom GPTs.
Verdict
Best for: Business owners who need real-time research capability alongside writing and analysis, or who want to build a custom GPT for their team.
If you're choosing between Claude Pro and ChatGPT Plus: use Claude as your primary writing tool and ChatGPT when you need current information or image generation. If you can only afford one, the choice comes down to whether you do more writing (Claude) or more research and varied tasks (ChatGPT).
Rating: 8/10 — slightly behind Claude for writing quality, ahead for research and breadth.
Tool 3: Perplexity
Category: AI-powered research and search Tested for: Competitor research, market analysis, industry news, fact-checking, answering specific questions with cited sources
What it actually is
Perplexity is an AI search engine. Rather than giving you a list of links to click through, it reads those sources and synthesises a direct answer — with citations so you can verify the sources and read further if needed.
It's not a writing assistant. It's not a content creator. It's a research tool, and the comparison isn't to Claude or ChatGPT but to Google.
What it does well
Research tasks are significantly faster than traditional search. When you need to answer a specific business question — "what's the average gross margin for a café in the UK?", "what are the main GDPR requirements for small businesses storing customer email addresses?", "what's the competitive landscape for independent bookshops in 2026?" — Perplexity produces a direct, sourced answer in thirty seconds. The same research done via Google requires clicking through multiple pages, reading past the SEO padding, and synthesising the information yourself.
Citations are included and clickable. Unlike AI tools that produce confident answers with no verifiable sources, Perplexity shows you exactly where the information came from. For business decisions where accuracy matters, this is important — you can check the source before acting on the information.
Follows up well. Perplexity handles multi-step research conversations well. "Tell me about X" followed by "now compare that to Y" followed by "what should I do differently given Z" — the context carries across the conversation in a way that makes research feel more like a conversation with a knowledgeable colleague than a series of search queries.
Where it falls short
Not a writing or creation tool. Perplexity can summarise and synthesise but it's not designed to produce polished finished content. Asking it to write a blog post or an email produces functional but mediocre output — this isn't what it's for.
Depth on niche topics is limited. For highly specific or localised questions — "what are typical rates for a mobile dog groomer in suburban Melbourne?" — Perplexity's sources are sometimes thin or not perfectly relevant. Its strength is on topics with a good volume of published, indexed content.
The free tier has query limits that active researchers hit within a day of intensive use.
Pricing
Free tier: Limited queries per day, access to the standard model. Perplexity Pro: $20/month — unlimited queries, access to more powerful models, file upload for document analysis.
Verdict
Best for: Any business owner who does regular research — competitive analysis, market research, industry news, regulatory questions, supplier research.
The free tier is genuinely useful for occasional research tasks. The Pro upgrade makes sense if you use it as a primary research tool multiple times per week.
Honest take: Most small business owners don't need to pay for Perplexity. The free tier covers 80% of research needs. If you find yourself using it daily and hitting limits regularly, the $20/month upgrade is straightforward value. If you're choosing between Perplexity Pro and Claude Pro, Claude Pro has more versatility for the same price — unless research is specifically your bottleneck.
Rating: 8/10 — the best research tool tested by a significant margin. Narrow use case but excellent within it.
Tool 4: Tidio
Category: AI chatbot and live chat for small business websites Tested for: Building and deploying a customer service chatbot, FAQ training, escalation flows, live chat management
What it actually is
Tidio is a customer service platform that combines live chat (for when you're available) with an AI chatbot called Lyro (for when you're not). It sits in the corner of your website as a chat widget and handles incoming customer questions automatically or routes them to you when human judgment is needed.
It's covered in more detail in Article 2 of this series. This review focuses on the comparative experience of actually using it versus alternatives.
What it does well
Setup is genuinely as fast as advertised. Installing the widget, configuring basic availability hours, and loading FAQ content can be done in two to three hours for a simple business. This isn't true of all chatbot platforms — some require significantly more technical investment before anything useful is running.
Lyro's AI responses are good on straightforward questions. When trained properly with well-written FAQ content, Lyro handles routine questions accurately and in a tone that doesn't feel robotic. The key — as covered in Article 2 — is the quality of the FAQ content you feed it. Garbage in, garbage out, but with good input the responses are genuinely useful.
The live chat and AI chat work well together. The handoff from bot to human is smooth — the conversation history carries over so you don't start the human conversation from scratch. This matters more than most people realise. The most frustrating customer service experience is explaining your problem twice.
Integrations are solid. Tidio connects to Shopify, WordPress, Wix, and most major website platforms with no code. It also integrates with email and CRM tools for lead capture.
Where it falls short
The free tier's AI conversation limit (50/month) runs out quickly for any business with meaningful website traffic. Fifty AI conversations is less than two per day — for a business getting regular website enquiries, you'll hit this within two weeks.
Lyro struggles with complex or multi-part questions. When a customer's question involves nuance, context, or multiple issues at once, Lyro often gives a partial answer or an incorrect answer stated with false confidence. The escalation to human works, but not always fast enough to prevent a frustrated customer.
Reporting is basic on lower-tier plans. Understanding what questions your bot is handling, where it's failing, and which conversations are converting into enquiries requires the higher-tier plans. On the free or entry-level plan, the visibility into bot performance is limited.
Pricing
Free plan: Basic live chat, Lyro AI limited to 50 conversations/month. Starter: $29/month — more Lyro conversations (200/month), basic automation. Growth: $59/month — up to 1,000 Lyro conversations, more detailed analytics.
Verdict
Best for: Service businesses, retailers, and any business with a website that gets regular customer enquiries and where response time matters.
The free tier is worth installing just to see how many website visitors are trying to chat with you. Many business owners are surprised by the volume. Once you've confirmed there's genuine demand, the $29/month upgrade pays for itself quickly if it's converting even one or two additional enquiries per week.
Not recommended for: Very small businesses with minimal website traffic, or businesses where all customer interaction happens in person or by phone rather than online.
Rating: 7.5/10 — strong within its use case, held back by the aggressive limits on lower-tier plans.
Tool 5: Apollo.io
Category: B2B lead generation, prospecting, and outreach automation Tested for: Building prospect lists, email finding, personalised outreach sequences, tracking open and reply rates
What it actually is
Apollo.io is a sales intelligence and outreach platform. It has a database of over 200 million contacts and 60 million companies. You filter by industry, company size, job title, location, and a range of other criteria to build prospect lists. Then you find email addresses, write outreach sequences, and send them — all within the same platform.
It's most relevant for B2B businesses — consultants, agencies, service providers, suppliers, or any business whose customers are other businesses rather than individual consumers.
What it does well
The database is extensive and the filtering is powerful. Building a list of, say, 100 independent restaurants in a specific city with a turnover over a certain threshold, filtered to show only those where a decision-maker has an active LinkedIn profile, takes about ten minutes. The quality and depth of the data available at the free tier is genuinely remarkable — this level of prospecting intelligence cost thousands of dollars per month five years ago.
Email finding is reliable. Hunter.io is the traditional standard for finding business email addresses. Apollo's email finding is comparable in accuracy and comes bundled with the contact database rather than as a separate tool and subscription.
Sequence building is functional and accessible. You can build a three or four-email outreach sequence — the structure covered in Article 4 — directly in Apollo, set the timing, and have it send automatically to your prospect list. The interface is more accessible than dedicated email outreach tools like Salesloft or Outreach, which are enterprise products not designed for small businesses.
The free tier is genuinely generous. 10,000 export credits per month covers a significant amount of prospecting activity for most small B2B businesses. This is not a crippled free tier designed to frustrate you into upgrading — it's a functional product at zero cost.
Where it falls short
Email deliverability requires attention. Sending cold outreach at any scale requires proper domain warming, SPF/DKIM authentication, and careful management of sending volume. Apollo provides the infrastructure but doesn't prevent you from making the mistakes (sending too much too fast, using your main business domain, ignoring bounce rates) that get your emails marked as spam and your domain reputation damaged. If you're new to cold email, read up on deliverability basics before running any sequence.
The AI personalisation features are useful but not magic. Apollo has built-in AI that attempts to personalise outreach based on a contact's LinkedIn profile or company news. The results are inconsistent — sometimes impressively specific, sometimes generic in a way that's more embarrassing than no personalisation at all. The manual personalisation approach from Article 4 (researching each prospect briefly and writing a specific first line) produces better results than Apollo's automated personalisation for a small, targeted list.
Data accuracy degrades on less prominent contacts. Apollo's data is excellent for well-indexed companies and senior decision-makers. For smaller businesses, junior contacts, or companies with limited online presence, the email addresses and contact details are sometimes outdated or incorrect. Verify before sending.
Pricing
Free tier: 10,000 export credits/month, basic sequences, email finding. Basic: $59/month — higher limits, more sequence steps, better analytics, more AI features. Professional: $99/month — advanced features including intent data and more AI personalisation.
Verdict
Best for: B2B service businesses, consultants, agencies, suppliers — anyone whose growth depends on finding and reaching decision-makers at other businesses.
Start with the free tier and use it for one month before deciding whether to upgrade. If you're building lists regularly and running multiple simultaneous sequences, the Basic plan at $59/month is worth it. If you're doing occasional targeted outreach (one campaign per month, under 100 prospects), the free tier is sufficient indefinitely.
Not recommended for: B2C businesses, local businesses where customers find you rather than you find them, or anyone without basic email deliverability knowledge — the risk of damaging your domain reputation is real.
Rating: 8.5/10 — the most powerful free tier of any tool tested. Held back slightly by the deliverability learning curve and inconsistent AI personalisation.
The head-to-head comparison
Here's a straight side-by-side for the two decisions most business owners face: which AI writing assistant to use, and which specialist tool to prioritise first.
Writing assistant: Claude vs ChatGPT
| Claude | ChatGPT | |
|---|---|---|
| Writing quality | Excellent | Very good |
| Real-time web access | Limited | Yes (paid) |
| Image generation | No | Yes (paid) |
| Custom configurations | Projects (paid) | Custom GPTs (paid) |
| Free tier | Good, daily limits | Good, daily limits |
| Paid tier | $20/month | $20/month |
| Best for | Writing, proposals, analysis | Research, varied tasks, team tools |
Recommendation: If you do one thing primarily — write — start with Claude. If you need a Swiss Army knife that does research, writing, and images, ChatGPT is slightly more versatile.
Which specialist tool first?
| Business type | First specialist tool | Why |
|---|---|---|
| Any business with a website | Tidio (free) | See how many customers are trying to reach you |
| B2B service business | Apollo.io (free) | Build your first prospect list this week |
| Retail / e-commerce | Klaviyo (free) | Start building post-purchase email flows |
| Trades / field service | Jobber (free trial) | Get off paper-based job management |
| Food / hospitality | Deputy or 7shifts (free) | Automate staff scheduling |
| Any business | Perplexity (free) | Faster research for any business question |
The one tool to start with if you're new to all of this
If you currently use no AI tools and you've read this far without knowing which to start with, here's the answer: Claude's free tier, for writing tasks, for two weeks.
Not because it's definitively the best tool across all categories. But because:
- It's free to start
- Writing is the task with the fastest, most obvious return — you'll save time in the first session
- The learning curve is low — you describe what you want, you get it
- It builds the habit of AI-assisted work in a low-stakes way before you invest in specialist tools
Spend two weeks using Claude for every writing task you'd normally do manually. Count the hours saved. Then decide which specialist tool makes sense for your business type based on what's taking the most time after writing is handled.
That's the sequence that works. Not ten tools in one month. One tool, used consistently, with a clear measure of whether it's worth it.
A note on what changes
AI tools are moving faster than any other category of business software. The tool rankings in this review reflect testing conducted in early 2026. By the time you read this, features will have changed, pricing will have shifted, and new tools will have emerged that deserve a place on this list.
AInstein reviews this guide twice a year and updates it based on current testing. If something significant changes between updates, we cover it in the weekly briefing. That's the fastest way to stay current without spending your own time evaluating every new tool that launches.
Final ratings at a glance
| Tool | Category | Rating | Free tier | Paid from |
|---|---|---|---|---|
| Claude | AI writing assistant | 9/10 | Good | $20/month |
| ChatGPT | General AI assistant | 8/10 | Good | $20/month |
| Perplexity | AI research | 8/10 | Good | $20/month |
| Apollo.io | B2B lead generation | 8.5/10 | Excellent | $59/month |
| Tidio | Website chatbot | 7.5/10 | Limited | $29/month |
Want this review updated twice a year and delivered straight to your inbox — along with weekly coverage of the AI tools and tactics that are actually useful for running a small business? Subscribe to AInstein. No sponsored content, no affiliate links in the editorial. Just the honest version.
You've reached the end of the AInstein Foundations series — ten guides covering every major AI use case for small business owners. Start with the article most relevant to your biggest current problem, and work through the others as you go. The full series index is here.
Get the week's most important AI developments for business owners — every Monday morning, free.