Running Before You Walk

The marketing hype at the moment is intense. The various platforms are promising that you can just type a few words and the outcome you want will just magically appear. But if you’ve been working with agentic AI, you know that the words have to be really specific, and the outcomes are variable. Of course, AI does open up automation of tasks previously not possible, such as:

• Responding to customer enquiries
• Prioritizing and drafting email
• Extracting data from messy invoices
• Summarizing legal documents

And I get the appeal of wanting to keep up with what’s new. I also know that at the end of the day, the only thing that matters is better business outcomes. AI is good, but it’s expensive, and I’m seeing a lot of people ignore the business tasks that can be automated with traditional tools, and could have been for years, in favour of the latest trend. It’s like buying a £450 pair of running shoes to run the London Marathon before you’ve mastered running around the block. You might look flash at the starting line, but you’re still going to struggle if the foundational fitness isn’t there.

Back to basics

While everyone is busy trying to build AI agents to manage their calendars, there is a massive amount of money being left on the table by ignoring “boring” traditional automation. Traditional automation is deterministic. If this, then that. It requires zero creative thinking from a machine, which means if it works once, it works every time.

Traditional automations such as:

• CRM to Accounting Sync: When a deal is marked as Closed Won in your CRM, does it automatically generate the invoice in Xero or QuickBooks? Or is someone still copy-pasting client details across tabs?
• Instant Lead Routing: When a high-value prospect fills out a form on your website, does your sales team get an instant Slack or Teams notification with the core details, or does it sit in an unread inbox for 24 hours?
• Centralized Onboarding: When a new client signs a contract, does your system automatically create a new project folder in Google Drive, a new board in your project management tool, and send a standard welcome email?

Predictable vs Probabilistic

AI is probabilistic. It uses complicated maths to guess the next best word or action based on patterns. That makes it brilliant at creative tasks and handling unpredictable human input. But do you want your financial reporting to be probabilistic? Do you want your client data onboarding to be a “best guess”? If a task requires absolute accuracy and follows a strict set of business rules, use traditional automation. If a task requires interpretation, nuance, or language translation, that’s where AI shines.

Don’t get me wrong - we use AI at Tessellium, and we love what it can do. It has enabled automations that were impossible a few years ago. But we use AI to enhance solid workflows, not to fix broken ones. Trying to build an AI layer on top of a fragmented, manual data structure will only multiply the problems.

By cleaning up the basics first, you get immediate business wins for less spend, and you build the digital infrastructure required to actually deploy advanced AI successfully later down the line.

I don’t know about you, but I haven’t met many high performing people who haven’t nailed the simple stuff.

Need help on a fundamental level? Talk to us about reinforcing your business foundation.