Transaction risk intelligence
Automated supplier statement reconciliation
Supplier risk intelligence
Managed AP recovery audit
Supplier file performance review
Forensic risk analysis
Consulting and bespoke reporting
What’s your AI fluency?
Your team are further along your AI journey than many vendors give you credit for. In a previous blog, we talked about your AI fluency ladder – and many teams are around that third rung of the ladder. You’re likely already using technology with AI built into it, whether that’s specifically made agents in your ERP, or AI assistants approved by your IT team.
But the imperative for many businesses is to incorporate more AI and increase team effectiveness even more. So you may be searching for more advanced AI technology, and the fourth rung of the ladder – the next step – is to use AI-Native technology and get connected with your workflows.
But what is AI native software exactly?
It’s a software that has been built from the very start with AI functionality at it’s core. Most of these will have sprung up over the last few years with the boom of AI.
Others, like FISCAL, have been working with AI for a very long time, honing our algorithms with years of data and expertise to create a highly specialised risk analysis functionality.
The question is, how do you figure out who can give you the deep expertise and specialised functionality you are looking for?
The five questions to ask your AI-native supplier
When you’re discussing AI software with any AI company (be they AI-native or otherwise), ask these five questions to really get to grips with how their AI functions.
- How do you train your AI and how deep is your training data?
AI that isn’t thoroughly trained makes guesses. Depending on what the software does, this could cause more problems than it solves. However, if the supplier has a large quantity of labelled, domain-specific AP data behind it, it will create fewer false positives, and you will be able to trust the outcomes more. Ask specifically if their data spans your particular use case, for example, your sector, size, number of transactions and suppliers, and any other relevant information. - Can you explain how the AI makes the decisions it does?
“The AI said so” won’t hold up with auditors or the board – the software needs to identify the logic it has used for you so it can be referenced. It’s also a really useful way to find out how flexible the software truly is and whether you can set out your own rules and boundaries for it. - Is there any way for humans to intervene if the AI isn’t working properly?
Being able to create human checkpoints to validate some of the decisions, and potentially stop the AI from a certain path, is ideal for control. Finance is a hugely risky area – any good finance software needs to have controls like these built in. - Does the software connect with other technologies or AI?
This one is optional, but it’s extremely useful to know about. Open architecture and connectivity open up a new world of workflows. It’s the difference between operating the software on its own and enabling the team to integrate it seamlessly into your existing processes. - How does it integrate and what security measures are in place?
What you’re looking for here is audit-grade security and governance, especially around any AI use. For example, are they ISO 27001 compliant? Does your data touch an external AI at any point? Where is the data stored and where does data get processed? How stable are the integration methods used and how stable is the login method?
Bonus questions!
If you truly want to get into the detail to be sure of what you’re getting with their AI, ask for specific information including:
- How long have they used AI in their product and how often has it been updated or upgraded? Can they detail those upgrades?
- Where and how do they use AI?
- Are there chatbots/ agents or is the AI part of the core processing capability? Or is it both?
Can it speak to other technologies?
Lets go off-piste for a second and take a closer look at question number 4.
Most P2P vendors have built their own AI, their own agents, their own dashboards — and there’s a logic to it. They want to keep you using their software.
But we know that you issue is that you don’t have just one piece of software to do your work inside – there are different ones for different functions. And not having them connected means you end up flitting between all of them.
What’s especially jarring is if your team is working from an ERP or AI assistant primarily, but it isn’t linked up.
And here’s where software featuring and MCP connector is very useful. It lets your existing AI assistant securely query data from connected software. No extra login. No switching tools. You just ask your AI a question and it pulls a live, accurate answer.
It’s incredibly helpful if some of your existing workflows already run through AI, and for those team members who want fast access to real-time reports or data. Perhaps you’ll need a few quick figures on payment practices, which 10 suppliers are riskiest, a visualisation of spend trends or duplicate invoices. All of these queries are answered in seconds.
We’ve seen the difference it makes at FISCAL – we use an MCP connector to retrieve our own anonymised benchmarking data, informing us, so we can help our customers better.
So the question here is really whether the software you are investigating plays well with others, and allows you this heightened level of flexibility.
Most don’t ask this because they don’t know it’s even possible, but for the teams furthest up the AI fluency ladder, it’s an expectation.
The five questions to ask your AI native technology supplier
How do you train your AI and how deep is your training data?
Can you explain how the AI makes the decisions it does?
Is there any way for humans to intervene if the AI isn’t working properly?
Does the software connect with other technologies or AI?
How does it integrate and what security measures are in place?
Bonus questions specific to AI:
- How how long have you used AI in your product?
- How often is your AI updated or upgraded?
- Where and how do you use AI in the software?
- Are there chatbots or agents, is the AI part of the core processing capability, or do you feature both?
