What to Look for When Onboarding an AI System
Understanding the Challenges of Buying AI Systems
Buying AI systems brings new transformative capabilities to companies, introducing them, but only if appropriate systems are bought. With so many companies selling AI systems – from startups to well-established companies – what should the discerning buyer look for when onboarding a new AI system?
Training Time and the True ROI of AI
For many organisations, the promise of AI is the promise to repurpose human time. Whether that’s enabling users to focus their time on higher-value activities, or simply to reduce cost by no longer contracting tasks to humans which are highly automatable by machines, often the Return on Investment (RoI) calculation is focused on time-saving.
This ROI can be particularly lucrative when the AI system is replacing tasks previously conducted by highly skilled or professional workers. Generally speaking, the more specialist the human skill, the higher the hourly/ daily rate – and therefore the greater saving per hour saved by performing tasks in an automated manner by machine.
However, where this logic breaks down is when the AI system offering a time-saving solution itself requires large amounts of human time to train for the purpose in mind. Just like how fresh-faced graduates might have a high level of aptitude for a particular career but require years of experience before achieving peak economic output, the same is true for many AI systems.
Often, such blank-canvas systems require laborious work to tag particular data items in such a way as to train the system ‘what good looks like’. While a Proof of Concept (PoC) is often devised in order to show potential time savings of a live environment, such an environment is often staged in such a way as to help calculate an RoI but without delivering any value in the process.
That’s why at Easy Autofill (EA), we don’t offer customers PoC to help them evaluate our platform. Instead, we seek to prove value by offering the full software in a guided trial – so not only can the outputs of EA be demonstrated on customers’ actual data, but the whole customer experience can be evaluated at the same time.
If the Poof of Value (PoV) activity is a success, then prospects go on to become customers, but if not, then we just delete their data and stay friends. That’s why EA has much higher than average conversion rates from the pilot (PoV) stage to paying customers than the market average.
Of course, a key reason for this is how we demonstrate how little training is required for EA to work. In fact, all the training that EA requires is provided through usage – and so as that usage expands – so EA improves – but from the very first hour of usage, the experience is designed to deliver value to customers.
Why Purpose-Built AI Outperforms Generic Chatbots
The next key area to look for is to differentiate between generic solutions and those purpose-built for a particular use case.
In a world where almost every tech company has a Chatbot offering, it’s easy to get confused about what capabilities are available from AI. Unlike Chatbots like Co-pilot, which give you the ability to hold a conversation with AI about a particular document or set of documents, EA is an ‘agentic’ AI platform that performs tasks for our customers.
In the case of EA, our superpower is filling in forms, and uniquely, we do so back to the original format, whether that be PDF, Word, or Microsoft Excel.
Understanding the process of B2B form-filling, we understand that this means that the process by which a form is completed the old-fashioned way starts with the production of a first draft, and then moves to a review, edit, and approval process, before final submission. Understanding this means we’ve built an AI platform that uniquely solves each of these three stages.
A first draft answer is provided for each of the questions on the form with minimal fuss, and much quicker than a human could ever have achieved. Next, the workflow around answer review, edit, and approval is streamlined – that takes into consideration all the requirements we’ve seen from customers, such as word/ character count, and four-eye/ six-eye approval for customers in regulated industries.
Finally, and uniquely to EA, the answers are returned into the original file – perfect for wowing your customer by your ability to turn around their diligence request quickly, accurately, and efficiently – as opposed to the multi-week exercise which would have been conducted had it been done manually.
The Importance of Human Support in AI Adoption
AI systems are fallible, but too often large tech companies push new capabilities to customers, expecting them to not only buy and use their products, but also to provide their own customer support when things go wrong.
Such self-service support can be infuriating as well as costly in human time. As a business that’s entirely set up to save our customers’ time, we instead offer every customer one of our senior data analysts to step in whenever they need help, with no cost. We never charge for our time.
The reason we do this is simple. If ever we need to work on a particular task manually ourselves for a customer, we break down the steps involved and build each one into our engineering queue.
While modern AI capability is truly marvellous, some tasks are still beyond the ability of a machine to reliably automate time and time again – and it’s important for us to experience these limits ourselves rather than for customers to do so and for their experience to fall short of expectations.
However, those many tasks which are fully automatable – we thrive on working through manually, as this gives us the inspiration (and the incentive) to complete end-to-end.
If you’d like to see this in action, click the ‘Get Help’ button in the bottom-left corner of the screen when you’re logged into EA and ask one of our analyst team to step in and help you. If they can, they’ll get the AI to complete the task you need, and if not, they’ll likely complete it for you. If they don’t understand what you need, they’ll book a call with you to understand your requirements in greater depth.
Conclusion
Hopefully, this short guide gives some important practical things to look for when buying an AI system.
It’s important to find AI that’s not going to require vast amounts of training, as while it might get you eventually to a place where you are experiencing an efficiency gain, the burden and cost to get you there might be far greater than originally expected.
Secondly, generic AI platforms are great in their ability to adapt to many use-cases, but rarely are complete alternatives to conducting work manually – their flexibility becomes a limitation that itself creates more work.
And finally, look for AI that comes with human support as standard – not as an expensive afterthought. Our belief is that AI should save our customers time, but if that time saved is instead spent training systems, doing new tasks created because of AI, or doing lots of trial-and-error to get results, then what is the point of it?
Have a question? Book a quick call to discuss, or you can now try EA for free to feel it.