Let’s be honest: Many AI wins are small. For example if you:
- Extract some data points from a document,
- Let an AI check compliance of a contract clause with a playbook, or
- Translate legal texts from foreign jurisdictions for a first impression,
then on each instance, you just save a tiny bit of time, like a minute or two. As a consequence, often it will not be worth the effort opening an AI tool, transferring a document there, waiting for the AI to process and then checking the results.
Hence, to realise these “Piggybank AI wins”, you need to make sure they are easily available without relevant overhead. This usually involves building a digital workflow. The good news is: These Piggybank AI wins often add up quickly to considerate savings in time and costs. On top of that, they free you from exactly the kind of dull and annoying work you want to avoid anyway.
So to collect Piggybank AI wins, it helps having in mind the following:
Take a look at the whole process
Piggybank AI wins are sensitive to their context. Hence, take a process-view and look how people in your company (should) work together. This has two major advantages: First, checking each step of the process on potential AI improvements forces you to do dozens or even hundreds of micro-AI-evaluations. This gives you better chances at finding many of those small AI wins. At least compared to hoping to spot them in your everyday work by coincidence.
Second, it allows you to integrate Piggybank AI wins into workflows and to trigger them automatically at the right time. This way, you take the waiting time out of the process while avoiding changes of media, resulting in quicker execution and reduced error rates.
Involve manual checks
AI is great, but it can hallucinate. Hence, you better ensure every result produced by AI is checked by a human. At times, this can be done implicitly, e.g. when an AI categorises a document and triggers the appropriate process, but humans later on are involved and can revert the AI’s decision.
Still, there might be exceptions when human oversight can be omitted: If volume is high, costs per review are high, but the risks involved are low, one might decide to go with AI-only. Plus, it makes sense to compare AI results to a realistic human benchmark: Also humans are not perfect at doing tasks. So if an AI does a certain task on human level or better, you have the option to either save human labor, or combine human and AI review to reduce error rates.
Use no-code tools
Piggybank AI wins usually are collected in a digital workflow. To speed up the process of building these workflows, no-code often should be your choice, as it delivers certain kinds of software at a fraction of time and costs of custom software development. Plus, no-code lets you easily combine AI and non-AI wins: For instance, in document generation, populating templates often is more reliable than using AI. And ultimately, your goal is not to implement AI, but to speed up processes the best way possible.
Hang Loose!
Don’t be overambitious: Not every tiny AI win needs to be realised. When a task can be automated, but is not done multiple times a week, it probably is best to keep things as they are.
To wrap up, small AI wins can add up to huge time savings if included in a digital workflow. To achieve this, look at the whole process and ideally use no-code tools to speed up development and combine AI and non-AI wins.
