Artificial Intelligence (AI) has become a hot topic of conversation amongst friends and colleagues. And there’s been a sense of urgency felt among business leaders to understand how it’s changing how we use technology to run our businesses.
In fact, it has been the centre of align.me’s workplace for the past few months. It even inspired our ‘AI Madness Month’, where we invited staff to spend a few weeks trialling and testing all the things AI, not just ChatGPT, can do for business. We recorded our findings in the hopes of discovering different ways we can leverage AI for our work.
It was an interesting month, to say the least. Many avenues were discovered, and at its end, we gathered to discuss our findings, draw conclusions on how AI can be used, and more importantly, if it should be used, and if so, in what capacity can we make it work for us?
This became the topic for this quarter’s referral lunch, where align.me’s business partners and referrers gathered to learn about the different ways we can utilise AI in business.
We’re happy to say that this was one of our biggest lunches which is a testament to AI’s significance in the landscape.
What is AI, and why all the fuss?
Our lunch started with a wonderful introduction from Director Brett Bonser, who shared an amazing story about how AI and technology helped a close family member during a tough medical period. This introduction reminded us of the power of technology and how it can be used to harness and uplift the work of humans and is a great parallel to how we should think about AI in life and business.
Founder and CEO Hugh Macfarlane then started his discussion with the Dartmouth Summer Research Project, which gathered some of the brightest minds in computing and cognitive science to discuss Artificial Intelligence. This event was considered a seminal event for AI as a brave new field. These bright minds concluded that complete AI was possible and would require a combination of techniques from several fields, including computer science, mathematics, and psychology.
That was 1956.
Seventy years on, ChatGPT was launched as a powerful, human-like, and free interface into the large learning models being ‘trained’ by OpenAI known as GPT-3 and more recently, GPT-4. ChatGPT has since gained rapid exposure and use – 13 million users and counting.
While acknowledging that some in the lunch group were far more expert in AI development than us, Hugh provided some context to serve as a basic framework for the discussion to level-set a bit.
- Large Language Learning Models like GPT-4 are essentially very large sets of data that are ‘trained’ by humans to be usable. Structure, context, synonyms, short-hand, acronyms are all a part of the training needed to turn words into meaning.
- ChatGPT – the human-like interface that has the world buzzing – is one interface but is not the only one. Other chat interfaces can and do reference these same models and are joined by voice, APIs that enable applications other than ChatGPT to interact with GPT and other models, and soon image as forms of input.
After a 70-year gestation, the impact of AI on business is only just becoming clear to us mortals, and we took the opportunity to have an in-depth discussion about it all – the good, the bad and the ugly.
With the benefit of a few months of steep learning and borrowing from more-informed others, we’ve identified three modes of using AI for Marketing and Sales:
- For questions that are new, infrequent, or ill-formed, chat provides a great interface for iterating until you get the answers you need.
- Questions that are clearer, especially those which are asked often, might find the iterative and bespoke nature of chat a little repetitious and inefficient. For those, templated questions work well. A more advanced version of templates is a table in a document or a spreadsheet connected via API’s into the learning models. Both of these approaches allow for standardisation and efficiency.
- For frequently asked and well-formed questions, it is incumbent on the application developers to bake AI into their applications.
The good
ChatGPT is not the first AI-driven chat interface, but nothing has come close to garnering popular opinion to the same degree. It is the beginning of something incredible. It is a prime example of the power of current technology which we should be utilising to expand our businesses.
As it stands, hundreds of AI tools can be used for anything from automating business processes, writing love letters, generating fake movies, proofing grammar, and anything in between. The published and undocumented uses of ChatGPT are broad and sometimes a little unique.
One of our participants shared that he has asked ChatGPT to re-write some emails to his colleagues in the US in American conversational style, which he claimed was ‘gold!’ This is a perfect example of how ChatGPT can be used to adjust communication from one style to another. It can also be used to save time and to ensure proper grammar and flow – things can be missed if you’re busy juggling a hundred other tasks.
So, what can AI deliver you for now (or soon)?
Our first discussion of the day saw our leaders break into six groups to share their experiences with AI and what possibilities they see for their own sales and marketing journeys. The question posed was, “what possibilities do you see for your sales and marketing for the next few months, and which of those do you think are within your skills?”
Hugh tabled some of the more accessible and useful cases for ChatGPT specifically:
- Ideating
- Generating whole copy
- Expanding – pattern extension
- Coding / script development
- Proofing
And that as a provider of outsourced B2B marketing services, we felt an obligation to iteratively learn about and operationalise best-practice use cases across:
- Copy
- Design
- Web development
- Automation
- SEO
- Briefing & reviewing
- Learning
- And our own Funnel Plan™ application
The workshops identified some generalisable uses:
- AI applications can allow basic sentiment analysis without reading thousands of documents;
- AI can be used to create text posts for social media and blogs and is particularly useful for technical work (at least for creating drafts);
- Automation of spreadsheets and presentations to save time on menial tasks;
- AI can help businesses create copy for a personalised outreach that will help them tailor / customise at scale;
- Using it for script development (like macros) and
- Blog topic ideation.
The bad
While there are tremendous benefits to using AI in programs such as ChatGPT, there are many ‘bads’ as well. AI skeptics will be the first to say, “I told you so!” and their doubts are not misplaced. Our next discussion focused on a few of the present gaps in AI. Much like anything, AI is not perfect (and our Madness Month can attest), but nor does it claim to be.
There can be significant time lost trying to identify the correct actions for AI to deliver the desired results. In addition, depending on the task, there will be a need to spend time fact-checking. Hugh accused ChatGPT of ‘mansplaining’ (confident answers that lack facts).
There is also a limit to what ChatGPT can produce, with it only being able to access data up to September 2021, meaning the further we move away from that date, the greater the risk of information being out of date.
What should we NOT be doing with AI (for now)?
We asked our groups to consider the impact on their brand as well as jobs, including the quality of work and time wasted, in service of identifying and prioritising the things they want to deliberately NOT do.
Discussions identified several downfalls to using AI, including:
- Spending too much time to make it work properly and knowing what to input to get the best results will take some time to get used to.
- Sometimes, AI language doesn’t match the audience (US English vs UK English as an example), and more qualifying inputs will need to be submitted to make a simple post work.
- Over-reliance on AI creates a risk of copy which has us all sounding the same.
- Unreliable data, quotes, or statistics. It may try, but you will need to go and fact-check it, meaning more time is wasted.
The biggest ‘bad’ of AI we were all concerned about is unclarity about ownership and responsibility. The suggested ownership sits with the person who commissioned the asset, but there are a lot of grey areas in responsibility. For example, who would really be at fault if AI-generated content led to an accident or error? Can you sue OpenAI for libel or slander? Perhaps it’s best to avoid using AI for any matters subject to legislative or compliance obligations.
The groups felt strongly that we need to ensure that we don’t lose faith in humanity. Nor should we lose our ability to think critically and creatively. AI should never be used to replace humans; it hasn’t been implemented to reduce the number of jobs. Instead, it should be thought of as a tool to complement the efforts of human beings.
Other things to consider:
- Where does AI fit in, and what is its purpose to you?
- AI is a tool and should be treated that way; it requires intuition and context to make it work – a job for humans.
- A means or an end? Some users may be unable to differentiate between using AI as an aid and using it just for the sake of it.
- Technology is meant to complement, not replace. There are risks of becoming lazy, unthoughtful, and becoming dependent on AI.
For fun, we asked ChatGPT what we shouldn’t use it for. It responded as follows:
- Decision making
- Personal Interaction
- Creative tasks
- Ethical decision making
- Tasks that require physical dexterity (oh, right! Good to know)
The ugly
The biggest worries that have come from using AI such as ChatGPT is the impact of inadvertent misuse or misdirection. Consider data protection, intellectual property (do staff inadvertently share ‘secrets’ when they ask questions?), misleading content, and bias and discrimination.
Bigger concerns – more existential – like misalignment of goals between an AI and its creators are also in discussion. These concerns have piqued the attention of Tesla giant Elon Musk, alongside a group of AI experts and industry executives, who have called for a six-month pause in developing language models that are more advanced still than GPT-4, citing potential risk to society. Musk has expressed frustrations over ungated development and says that the safety concerns need to be properly studied and mitigated, with calls to regulators and technology leaders to take the time to consider the implications of self-improving, self-determine intelligence forms that might have goals that are not aligned to our own.
These are concerns for the cognoscente amongst industry heavyweights to deal with. And for those of us trying to run businesses effectively? What guardrails do we need to employ?
What risks do you see… and what guardrails do you need for each?
At the end of our day’s discussion, we focused on pinpointing the limits that we must establish for ourselves around AI’s ethical and sustainable use in our businesses.
We asked our groups what are the biggest risks they can control, and what guardrails they should implement to manage these risks. There was strong agreement that we need to avoid large-scale mistakes. We need to break work down into parts and set guardrails for each part to ensure each section uses AI safely.
To find the guardrails, businesses need to:
- Make sure to run tests and learn from mistakes. Every business is different, and every user is different.
- Understand your business needs to help establish boundaries for decision-making.
- Remember not to get too reliant on the tool; believe in your humans.
- Be mindful of customer data, legislation, policy and protocol.
- Assess the risks – seek expert help, and treat it like any other employee. Ask yourself who can use it and what types of questions and languages it understands.
- Know your brand and reputation and always be wary of the validity of what AI produces; ensure there is a control and approval process.
- Be knowledgeable enough about the outcome you want AI to produce, be wary of misleading and deceptive conduct and always fact-check.
Setting up guardrails in business is important to ensure that you’re doing your best work and allowing for the independence of workers by ensuring their efforts stay aligned with the business’ values.
Forging the trail: where to next?
There are endless possibilities to experience with AI. However, in these early stages, it’s best to experiment and learn from trial and error rather than just adopt a wait-and-see approach. AI has already gained momentum. There was a mutual understanding amongst the leaders that they can’t simply sit back and see; they need their teams to get out there and use it for themselves.
It might feel like a difficult landscape to navigate, but AI can present many rewards if done correctly and ethically. The most important takeaway is not to lose faith in the humans nor their ability to think critically and creatively.
Moving forward, we all agreed to be simultaneously proactive and diligent as we weave through this landscape, establish authority and delegation, understand the risks, and set up our guardrails accordingly.
Time for your own madness month
This referral lunch was a huge success, partly due to the excitement and intrigue surrounding this topic, but mostly thanks to the calibre of the attendees. Our discussions clearly showed that modern business leaders are keen to be innovative and learn new technologies, and are happy to do so together.
We want to thank our valued clients and their partners for joining an insightful conversation.
These lunches provide a great space to share knowledge and create value in return. We look forward to these networking opportunities to thank those who continue referring us to their network.