As a business growth consultant working with multiple clients, I’ve been knee-deep in testing AI tools – both for my own practice and alongside clients who are excited to explore how technology can be leveraged to solve real business problems right now.

Here’s the real story of what we’ve tried, what’s working, and what we’re still figuring out.

The Strong Foundation: Building a Fit-for-Purpose Knowledge Base

The Business Challenge

First up: we tackled the challenge of scattered information. You know the problem – crucial details spread across documents, emails, and people’s heads. At the most basic level, consolidating this information in a business is critical to ensure accuracy and consistency throughout customer service, marketing materials, employee onboarding etc. The challenge in a rapidly evolving start up environment is sparing the (human) resources to get this knowledge base constructed.

Approach

Our solution? Gathering the relevant sources, providing a logical structure for the knowledge base and letting AI do the heavy lifting in getting a first draft constructed, before leveraging the limited time of human experts for validation of the detail.

Core Components

What’s in it:

• Product details and service information.
• Company policies and procedures.
• Industry insights on relevant topics.
• Expert knowledge from team members

This isn’t just a fancy filing system. The principle for us was to build an asset that can be leveraged without capacity limits, and one that can be easily updated as and when circumstances require it.

Practical Applications

We’re using this knowledge base to:

• Query it through natural language on specific topics that we need to fact check.
• Power limitless content creation, that remains accurate, and on brand.
• Build highly relevant customer education materials in multiple formats.
• Serve as the foundation for our customer facing virtual assistant.
• Create content for webinars.
• Respond in an agile manner to customer questions originating from multiple touchpoints in a format that is suitable for the platform.
• Support a range of business-as-usual marketing activities such as email marketing, social media content generation, etc.

Yes, it took time to build. But now? Tasks that used to take days take hours. More importantly, everything stays consistent and on-brand.

Custom AI Assistants: Making the Impossible Possible

Running a consulting business means juggling multiple clients, each with their own unique needs. My secret weapon? Custom AI assistants built with the paid version of ChatGPT and set to not use chats as part of model training. Each assistant is provided with context such as:

• The client’s business background
• Their goals and objectives
• Brand voice and positioning
• Target market details
• Product specifics

Some assistants handle specific tasks like crafting email campaigns or social media content. Others generate detailed assessment reports. The best part? Clients who’ve adopted these assistants are seeing their teams become more efficient and consistent in their work.

By building these custom assistants, it saves me hours in generating extensive prompts or storing unnecessary prompts in a separate library, not to talk about the mental capacity saved when switching context from one client to the other.

Taking on Video Production

Here’s a fun one: we needed educational videos for our customers in key points in the customer journey but couldn’t justify spending budget on getting this done through traditional video production processes. So we had a choice to make, forego the videos all together or explore what is possible with AI video platforms. After testing both Synthesia and HeyGen, we went with HeyGen for its suitability of avatars and features.

Our current process:

1. Draft video scripts (using our previously mentioned knowledge base).
2. Get experts to review the content.
3. Design a customised template leveraging our in-house design experts.
4. Spend some quality time to transform our script with into a video using the power of AI.

Our initial set of 6 videos is now easily expandable, and we can go from idea to finished video in 2-4 hours. Are they perfect? No. Do they get the job done? Yes – the viewing numbers suggest they absolutely do.

Getting Smart About Virtual Assistants

The Evolution of Customer Service

Virtual assistants have moved beyond basic chatbots to become valuable customer service tools. The key is in how they’re implemented and trained to handle real-world interactions effectively.

Implementation Learnings

In working with specific clients we learnt that virtual assistants need proper training – just like human team members. We’ve learned to:

• Provide it with a clear and factual knowledge base, and proper onboarding.
• Set clear industry-specific boundaries.
• Test with tricky real-world scenarios.
• Test expanded language capabilities where appropriate.
• Fine-tune based on actual customer interactions especially in the early days.

These assistants are not replacing human interactions, but they absolutely complement the experience and have in certain instances exhibited incredible patience in the most challenging customer interactions.

Research: The New Way

Staying informed and capturing insights effectively is crucial for making strategic business decisions. Our toolkit has evolved significantly:

Research Enhancement

Perplexity has revolutionised our approach to market research and trend analysis by:

• Nearly cutting research time in half while improving depth.
• Connecting dots across different industry sources.
• Providing more current insights than when using previous desk research methods.

Content Verification

Notebook LM has transformed how we validate and utilize information:

• Rigorous source checking capabilities.
• Efficient analysis and interrogation of multiple lengthy documents.
• Game-changing audio content analysis features.

What’s Next on Our Radar

While we’ve made progress, we know there’s so much more to explore. Right now, we’re excited about:

• Getting our AI tools to work together better.
• Automating more complex workflows.
• Expanding what our assistants can handle.

Here’s what we’ve learned: start with real business problems, not the technology. Test thoroughly. Build on what works. And most importantly? Stay focused on actual results, not just the cool factor of new tech.

Let’s Compare Notes

Every business’s journey with AI is different. I’d love to hear what you’re trying out. What’s working? What isn’t? The more we share our experiences, the faster we all learn.

Working on your own AI initiatives? Let’s connect and swap stories. The best insights often come from comparing notes with others in the trenches.