AI Interview - Nick Mescher & Dean Cornish

Head of Operations, Gems Holden sat down with our Field CTO, Dean Cornish and Strategic Advisor, Nick Mescher to chat about AI, business and where it can be leveraged successfully in 2025.

Dean, you’ve been working within AI for a significant amount of time and have seen a huge leap in terms of ability and application. Can you give me a quick rundown of your AI experience particularly the last 2 years?

Absolutely, I’ve been working with AI since 2017 and I’ve been following the development in terms of capability and opportunity closely. Even the very meaning of what it is to be considered “AI” is constantly being redefined and this is one of the reasons I continue to be excited about it. Up until July of this year, I was leading the AI practice at a large consultancy where I lead the team that tests AI implementations.  I’ve spoken at conferences about the journey of introducing AI throughout our company and today, I support our clients here at PTP in my role as Field CTO.

Personally, I am really excited about AI and its adoption into the mainstream vernacular.  My particular areas of interest sit around Data & AI and the safety of AI factoring in risk and change so businesses can successfully implement it in an ethical and sustainable manner.

Nick, you have been advising clients on the practical ways they can introduce AI into their business. Can you give me a couple of examples of where you’ve seen AI leveraged effectively over the past 12 months here in Melbourne? 

The most common usages that I see of AI at present are in analysing large amounts of data and providing synthesised summaries in easily consumed formats for different audiences, these could be bullet points, story style, dashboards or presentations and are easily targeted towards specific messaging that would typically have taken a trained workforce hours (maybe even days to produce). The key to this usage is to ensure the people reviewing the outputs are knowledgeable and experienced enough to validate the content before trusting it as fact.  During the repeated use of these processes, there will be learnings made that ensure the same mistake will not be made twice.

 Dean, after hearing an advert on the radio espousing an AI vacuum cleaner (true story) Can you explain what are the most common types of AI?

There are two fundamental types- generative and discriminative (AI) models. Generative models synthesise data where no data exists and they make calculated assumptions- if you will, to produce the output. Examples of these are chat-bots like ChatGPT and image or video generators. Non-generative models perform computations based on the input data, such as classification of images, translating text into another language or predicting events. Each of these have different use cases they are well suited to. Sometimes in marketing AI is used purely as “sales puffery” and the actual link to AI is tenuous (at best). AI perhaps may have been used on data collected from customers to determine what features are important, and not in the product itself a consumer actually buys. I’d recommend a healthy dose of scepticism when seeing “Contains AI” in marketing material unless they’re specific about its use.

Dean, what should business leaders and executive team be doing to prepare for AI and AGI in the near future?

All businesses in 2024 should be starting to organise and actively curate and manage their data. This also means “cleaning” the data and removing duplicates, incomplete records, invalid data and so on. It may also mean decorating the data with meta-data to make it more easily ingested at a point later down the line by AI models so the data can be more easily classified or grouped. Context is everything. It is the difference between data meaning one thing, or something completely different and for the moment, only a human can do this reliably. AGI may change this of course. The quality of data is key. Executives should also be mapping out all their processes and identifying suitable use cases for AI with a view that some are low hanging fruit and are low risk opportunities to start their AI journey.

 Nick, do you agree with Dean in terms of how we best prepare in the near future for AI?

I do but I suppose I take a more operational perspective in my role as a Strategic Advisor and Board Director. For me, AI is company-wide: from Boards through to the Executive and Management and personnel, and there needs to an appropriate approach to preparing that starts with policies and knowledge-sharing.  Boards need to ensure there are adequate policies in place and they are being followed to protect the company, especially as more and more people will be experimenting with AI either directly in the workplace or at home on company devices or using company logins.  These policies should be an accurate reflection of the risk appetite of the organisation but also put in place clear boundaries.

AI can be scary, exciting, challenging and innovative all at the same time, and a clear education (knowledge sharing) program should be planned that is pitched at the different levels of the organisation, so individuals are made aware of the opportunities it can provide, what the business is thinking and how employees can play a part in the evolution.  Whilst many see AI as a buzz-word or trend, there is a need to de-mystify for all staff.     

 Dean, what are the limitations that you’ve observed with current generations AI and how rapidly is this changing?   

The limitations presently are breadth and depth- namely in data and comprehension. For example; if you’re generating an image, there will be little tell-tale markers of AI involvement- look for the nonsensical- such as too many or too few fingers, or impossible things, such as the camera placement or laws of physics not being respected. Keep in mind generational shifts, in little under a year, some of the models have gone from producing utterly unrecognisable results to those that are almost accurate barring flaws that only an expert would detect. It won’t take many more generations for the technology to get to the point where it is indistinguishable from ‘real’. This isn’t as far away as we previously had thought.

 Nick, how can a SME use AI to help them grow and/or cost save?

AI is well-positioned to help SME’s both grow and to be more efficient in their work practices, leading to cost savings. At a base level AI can be used to automate and enhance repeated processes by constantly learning and teaching the most efficient path thru a chain of activities, this has been used very effectively to learn from past mistakes and trace back to the root cause of the problem.  In this way a potentially minor process or decision, early in a process life-cycle, can be altered (corrected), saving large amounts of re-work and time in the e2e process.  Many businesses have over-time, built inefficiencies into their work without recognising it, the focused use of AI can not only help eliminate these but also suggest new steps that may further increase the efficiency of the business.

Growth from AI can come from several sources, simplistically marketing and business development activities can be beneficiaries of increased knowledge leading to more pointed targeting of high potential customers, who have a need and it can then assist in defining the product offering to offer the best chance of winning the business.  Even more simplistically, the promotion of your business as being AI-aware (see response to question around Boards preparedness), will present an image of a business that is keeping current and market leading.      

Dean, do you agree with Nick’s argument around automation and time efficiencies that AI can bring to SME?

Absolutely, good opportunities for growth and savings for SME’s are identifying time consuming activities that they’ve done potentially thousands of times before and perform these activities with a generative AI solution, then use their deep domain and subject matter knowledge to improve the output and ensure it is accurate. The learning comes from exploring the best ways to achieve the most ideal output. Sometimes this means using the English language in a completely different way than it was intended.  To put it in context- prompt engineering is little more than using English to reverse engineer a “black box” (of unknown rules). As you discover the rules, you can change your language pattern to achieve specific results. It takes practice and learning and is a skill to be honed like any other. Someone who is effective with prompting can produce high quality results in moments compared to someone else who can be iterating for hours and still not be satisfied with the result.

Dean, what are some good practices to employ AI safely in your business?

First, have an AI policy that addresses how staff should use AI, especially around how to handle secrets and how to use the product of AI. Second, understand what AI your staff are already using and provide a safe, vetted, internal version that can be used instead and curb the use of external/public instances. You don’t want your customer’s or your company’s data appearing in the next public version of OpenAI’s GPT. Lastly (and arguably, most importantly), train your staff about what AI does and ideally how it does it. De-mystifying AI is a crucial part of understanding AI. Once it is no longer perceived as being “magic” you can start to identify suitable use cases from an informed viewpoint and the product you produce will be more ethical and likely to add value to your business and your customers.

Keen to talk more on AI? Why not give us a call today to chat to one of our team.