Broadly, Artificial Intelligence (AI) can be defined as the following:

1. The ability of a computer or other machine to perform those activities that are normally thought to require intelligence.

2. The branch of computer science concerned with the development of machines having this ability.

3. Intelligence exhibited by an artificial (non-natural, man-made) entity.

Whilst the media has always focused on #3 with Terminator etc, #2 is just for specialist coders so we are concerned with #1 – computers being able to perform tasks normally reserved for humans. AI is driven by Machine Learning (ML), where a computer processes available information to build models for prediction, and it is this that has been popularised by OpenAI’s Chat GPT. As anyone who has used Chat GPT knows, it will never replace a human, but can replace the most routine parts of tasks that we perform.

Chat GPT is a Large Language Model trained on trillions of pieces of information from the internet and elsewhere. Amazingly, when you type in a question it is ‘just’ evaluating the next best word to provide. Consider that the successors to Chat GPT will learn from an organisation’s entire dataset and provide answers to business questions, i.e. what open tasks we have where clients are most likely to raise issues.

We have been stuck in the ‘Productivity Gap’ for many years where the technology has increased but output per person has decreased. Currently we have a paradox where we are hugely short of skills but those in work are producing less. AI will improve this.

If you think of an AI as being the best assistant you ever had, automating your boring tasks, and helping you get to your valuable work, it’s clear that productivity is about to get a big boost. Goldman Sachs estimates that two thirds of US occupations will be improved by AI - I think this is too low, but on the flip side 60% of all workers do jobs that didn’t exist in 1940, so there’s a huge world of opportunity out there.

We have helped clients use AI since we started, but it’s been a niche activity for most people and organisations, firmly in the ‘too hard’ pile. The release of Chat GPT in December 2022 has changed all that in the same way that Google took internet search mainstream.

However, while Chat GPT can speed up many consumer tasks like writing emails, essays, recipes or reports its success means that it is often too busy to use, and then only for one question at a time which makes it impractical to use for business purposes. With many businesses using Alteryx to clean, sort, and analyse data, we as global experts felt best placed to build GPT connectors, and they are amazing and reliable.

In the Channel Islands we have one of the highest concentrations of Alteryx users in the world thanks to the early adoption of the technology, not just by financial and professional services but across industry, with telecoms, government, retail and even charities. Jersey alone has more than 1% of the world's advanced Alteryx users, and a growing number of businesses are using it to improve quality, simplicity, and repeatability in their regular processes. We therefore know there are many people who we could help get started with AI now, and indeed we’ve had well over 200 downloads already!

The Continuum ChatGPT Connector is an Alteryx Macro which hides the complexity of ChatGPT’s end points or APIs (Application Programming Interfaces) from the user. Because the Connector takes your requests through an API (the back end of ChatGPT) speed can be controlled so it always works no matter how busy the user side of ChatGPT appears.

The Alteryx user can then simply drop in their installed GPT macro, configure it with their user information, then connect it to their dataset. The Alteryx user builds a workflow as normal, combining as many of their different data sources as needed with transparent business logic to produce a repeatable, no code dataset and create a question, e.g. ‘Can someone aged <age> vote in <country>? Yes or No.’ When you drop in your installed GPT macro, the macro will output the Chat GPT answer for every row, in this case allowing you to filter out your clients who can’t vote.

The Edits Connector is for GPT v3 users and points to the Edits endpoint which helps correct an input string for spelling and grammar but is being superseded by the Completions Connector for later versions (v3.5 turbo and v4) which is where the fun stuff happens.

The Completions Connector asks a natural language question based on your data, much as you would in Chat GPT, and then receives the reply, allowing you to act on the answer in your workflow. This is where this is truly exciting because Alteryx allows you to combine data from all your systems into a question – what’s known as ‘prompt engineering’.

Imagine asking ‘On a scale of 1 to 10 how profitable is a customer who last year spent <spend> on goods with a cost of <cost> and on which we spent <number of hours> hours?’ People are already updating their LinkedIn profiles to be ‘prompt engineers’. No need, we’ve got Alteryx.

We created the Connector for the Alteryx community and have put it out there for all users for free, with no responsibility for its use nor guarantees of GPT accuracy. This means that no matter what the size your organisation if you have an Alteryx licence you can take advantage of the latest AI. If you don’t have Alteryx, we are of course delighted to help you with a free trial.

Amongst uses so far, we have seen sentiment analysis of customer feedback and emails.

Alteryx reads the messages, has GPT score the sentiment from 1 to 10, then where a score is below or above a certain threshold you can then generate a tailored email or choose to escalate the issue. This is usually the domain of specific machine learning platforms that take many hours to train for each language and use case.

Another client is using Alteryx to read in the text of PDF invoices before asking GPT to return only the Supplier, Invoice Number, Date and Amount in a table. And it works! This usually requires a specific system or coding to use cloud platforms.

Finally, we used our connector to win a UK hackathon against 20 other teams where the challenge was to identify the correct product code for any good or service in English, a task that usually takes an hour per code. Not only did GPT deliver the best of any team in 15 seconds, but it also works in any language and can return multiple different standard product codes.

The usages are as endless as in Chat GPT, and indeed we have much more control so we can tune the preferred response, for instance make it formal depending on how well we know the person.

Some organisations and indeed countries already have strict policies on using ChatGPT due to fears over data privacy. Alteryx allows you to remove or generalise sensitive data before asking the question. You can then re-join the results to your data in Alteryx to use the personal identifiers that never left your organisation.

Alteryx already allows business users to do an amazing amount with your business data without the need for code. However, using advanced AI like this opens a quantum leap in quality and technical simplicity.

More than that, Chat GPT is only just the start. It’s hard to believe but the AI will only get better from here, and the uses will expand as AI from Google, Amazon, Meta, and many others come online. We will be able to connect to all of these in the future too, allowing Alteryx users to pick the best AI for different purposes, bringing all the data back together in your own emails, automations, reports, or dashboard.

If you’ve already got Alteryx, we can help you connect our connector. If you don’t, we can come and do it for you. We’ve a team of experts and the technology has now reached the point where we can demonstrate clear value for any provable business case in a short period of time.

We are trusted to deliver by our 90 clients, including RBS, Global Radio (Capital, LBC, Classic FM), Standard Bank, Government of Jersey, and JTC with whom we have many testimonials.

We can take anonymised data sets for free and using AI prove whether there is a business case, and if so, implement the solution in weeks not months.

If you’re doing a system implementation or data migration, get us involved. Our team will help improve the quality, speed, and risk profile of your project immediately. If you aren’t but are stuck with too many projects and just don’t have time, you need to get your data right.

We can also come in and help put together repeatable, risk-based data quality dash boards that will get you into a better space ready for using AI to find areas of unexpected cost or opportunity.