I’ve wasted money on consultants I didn’t need. I’ve also tried to DIY things I had no business touching and wasted even more money. After running three startups and working with AI across all of them, I’ve got a pretty clear framework for when to get help and when to just figure it out yourself.

The short version: most founders outsource too early on the easy stuff and too late on the hard stuff. Here’s how I break it down.

Category 1: Just Do It Yourself

These are the projects where hiring a consultant is honestly overkill. You’ll learn faster by doing, and the tools have gotten good enough that you don’t need deep technical expertise.

Setting up a customer service chatbot. Tools like Intercom, Drift, or even a custom GPT can get you a working chatbot in an afternoon. We set ours up for about $50/month in tooling costs. It handles roughly 40% of inbound support queries now. A consultant would have charged us $5,000-$15,000 for essentially the same outcome.

Basic workflow automation. Connecting your CRM to your email marketing platform. Auto-generating reports from your database. Routing support tickets based on keywords. This is Zapier-and-Make territory. Total cost: a few hundred dollars per month and a weekend of your time. There are thousands of tutorials online. You’ll be fine.

Content and marketing AI. Using AI for drafting blog posts, generating social copy, summarising customer feedback — this is table stakes now. You don’t need a consultant to tell you how to use ChatGPT or Claude. Experiment, find what works, build your own prompts. Cost: whatever you’re paying for the subscription.

Internal knowledge bases. Tools like Notion AI or Slite can turn your messy internal docs into something searchable and useful. Again, it’s a weekend project, not a consulting engagement.

Category 2: Maybe Get Help

This is the grey zone. You could potentially DIY these, but the risk of getting it wrong is higher, and mistakes here actually cost real money.

Data pipeline architecture. If you’re connecting multiple data sources and building something you’ll rely on for decisions, it’s worth at least getting an expert to review your approach. I tried building our analytics pipeline myself once. It worked for three months, then fell apart spectacularly when we hit 10x our original data volume. A consultant’s review would have cost $3,000-$5,000. Rebuilding the pipeline cost us $25,000 and two months of engineering time.

Custom integrations between systems. Connecting your proprietary backend to a third-party AI service isn’t always straightforward. API documentation lies sometimes. Edge cases multiply. If the integration is mission-critical, at minimum get a technical review before you ship it.

AI-powered product features. If you’re building AI into your product — recommendations, personalisation, smart search — you can prototype yourself but should probably get expert eyes on it before you go to production. The difference between a demo and a production-ready feature is enormous, and customers notice.

For this middle category, my advice is: start the work yourself, then bring in a consultant to review, stress-test, and improve what you’ve built. It’s cheaper than starting from scratch with them, and you’ll learn more in the process.

Category 3: Definitely Hire Someone

No hedging here. These are the areas where DIY is genuinely risky.

Anything involving sensitive customer data. The moment you’re processing personal information, health data, financial records, or anything covered by the Privacy Act, you need someone who knows what they’re doing. A data breach or compliance failure will cost you infinitely more than a consultant’s fee. We’re talking $20,000-$80,000 for a proper implementation, but that’s insurance money as far as I’m concerned.

Custom ML model development. If off-the-shelf models don’t cut it and you need something trained on your specific data, hire someone. Building, training, validating, and deploying custom models is a specialised skill. I’ve seen founders burn through $100K trying to do this with a junior developer before giving up and hiring an ML engineer anyway.

Regulatory compliance and AI governance. Australia’s AI regulatory landscape is evolving fast. If your AI system makes decisions that affect people — hiring, lending, insurance, whatever — you need someone who understands the legal and ethical framework. This isn’t optional anymore. SmartCompany has been covering the regulatory shifts well if you want to stay updated.

Production-scale infrastructure. Going from “it works on my laptop” to “it serves 10,000 concurrent users reliably” is an engineering challenge that eats startups alive. Get help.

How to Find Good Help

When you do need a consultant, here’s what I look for. First, they should be able to explain what they’ll do in plain English. If they can’t, they either don’t understand it themselves or they’re trying to make it sound more complicated than it is. Both are bad.

Second, ask for references from companies your size. A consultant who’s brilliant with enterprise clients might be completely wrong for a 15-person startup. The scale, the budget, the pace — it’s all different.

If you’re based in Australia and looking for practical AI guidance, I’d recommend talking to AI project delivery firms like Team400 who actually work with startups and growth-stage companies. You want people who understand that your budget is measured in thousands, not millions.

The Real Framework

Here’s my honest decision tree: If you can find a tutorial that covers 80% of what you need, do it yourself. If you’re spending more than two weeks stuck on something, get a review. If someone’s personal data or your company’s legal exposure is involved, hire a professional from day one.

That’s it. No complicated matrix. No “it depends” waffle. Just an honest assessment of what you know, what you don’t, and what the downside looks like if you get it wrong.