10 Key Insights from our Event on ‘Generative AI: How to Make It Work for You’ in Collaboration with TechTO

Generative AI — if you’ve been paying attention, it’s everywhere. From boardrooms to tech conferences, it’s the only tech everyone’s talking about.

But for every flashy thing, there’s a quieter, less glamorous side.

You’ve heard the stories of massive budgets blown, hours spent, and at the end? A project that just didn’t quite deliver.

If you haven’t, here are some recent ones. (Source)

➡️ Air Canada pays damages for chatbot lies.

➡️ iTutor Group’s recruiting AI rejects applicants due to age (paid $365,000 to settle a suit)

➡️ Amazon AI-enabled recruitment tool only recommended men (and guess what, they scrapped the project)

This shows the gap between the ‘hype’ and ‘what actually works’ can be, well, HUGE!

But failure isn’t the end of the story. It’s just the beginning of figuring out how to get it right.

At Azilen Technologies, we’ve seen both sides of the AI coin, and that’s why we partnered with TechTO, Canada’s largest tech community, to host an event that cuts through the noise: “Generative AI: How to Make It Work for You.”

Azilen CEO, Naresh Prajapati, along with Saroop Bharwani from Senso.ai and Jennifer Arnold from Minerva AI, took the stage to share the stories you don’t usually hear – the near misses, the pivots, and, crucially, the strategies that actually got them across the finish line.

This blog covers the key takeaways from the event, stripping away the ‘hype’ and getting real about how businesses can ‘actually benefit’ from Generative AI today.

Picking between LLMs and SLMs depends on various factors, including –

✅ The specific tasks you aim to perform

✅ The availability of data

✅ Computational resources

✅ The trade-off between accuracy and efficiency

To achieve a balanced approach, start with LLMs for faster prototyping and gradually optimize with SLMs.

The below diagram highlights the parameters that organization need to consider.

Not every problem needs the firepower of an LLM. In fact, a lot of issues can be solved faster and cheaper with SLMs — Smaller Language Models.

They’re quicker to train, less resource-intensive, and a lot easier on the budget.

It’s really reshaping how we think about scalability in AI.

Sure, LLMs can do just about everything, but do you really need that for every task?

Task-specific SLMs are built for specific jobs, which makes them more efficient at solving targeted business problems.

For example, MathGPT is excel at solving complex mathematical problems, while CodeBERT shines when it comes to generating and understanding code.

In contrast, general-purpose models like ChatGPT or Gemini might give you a good answer, but they may lack that precision in specialized areas.

The chart presents the general-purpose LLMs and domain-specific SLMs.

LLMs are great, but they’re expensive — training and implementing them can burn through your budget fast.

SLMs, on the other hand, offer a much more affordable solution.

You get solid performance without the heavy price tag, which is a huge win for companies trying to be both cost-conscious and competitive.

Success in AI isn’t just about the data itself but how you stitch it together to create a compounding effect.

A deep funnel approach is critical — making sure your data ecosystem is interconnected to drive real, actionable outcomes.

It’s tempting to try and solve everything with AI, but that’s a mistake. You need to zero in on a single, mission-critical problem.

If you don’t have a clear understanding of what problem you’re solving, stop and rethink.

AI only works when you dig deep into the details, solving the actual root issue — not just scratching the surface.

Before going all in, it’s important to test your AI use case through a Proof of Concept (PoC).

Does it actually provide value to the organization? Is it functional and impactful? These are the questions that need answering before scaling up, ensuring you’re investing in something that’s going to drive real results.

Generative AI is incredibly powerful when it comes to processing huge volumes of data.

It can sift through overwhelming amounts of information, identifying patterns and providing insights that would take humans ages to uncover.

The key is using AI to turn that mountain of data into something actionable, something that can drive real business decisions.

Focusing on a niche market can be your best bet for early success.

By honing in on a specific sector, as Senso.ai did, you can tailor your offerings to meet distinct needs, making it easier to gain traction and build credibility.

In a competitive landscape, stagnation can spell disaster.

Continuous iteration not only helps refine your product but also keeps you responsive to market changes and emerging trends, ensuring you stay ahead of the curve.