Three tasks small businesses can automate with AI
Automation does not have to mean a big system. Often it is three small tasks eating an hour a day that could take ten minutes instead.
Read the article
Service
AI automation for the tasks that keep repeating
Most people asking about AI do not want a system. They want to stop doing a task. Usually the same task, every week, one that needs very little judgement but still has to be done by a person.
Splenix works on the dull, useful end of AI. Not grand plans about machines taking over, but three or four specific tasks that stop eating the whole morning.
We always start with one task. It runs alongside the way you do it today, so you can see whether it genuinely saves time before anything larger gets built around it. If the answer is no, you spent very little finding that out.
Who this is for
Businesses buried in similar enquiries
Most of the questions resemble each other. What does it cost, where are you based, could you take a job in May. The answers already exist, they just get written again every time.
Companies selling in two languages
All the content doubles, and in practice one language version falls behind until it no longer matches the other.
Teams moving data by hand
Somebody copies fields from an email into a spreadsheet, or from a spreadsheet into a business system. It repeats, it has clear rules, and mistakes are easy to spot.
Managers buried in long documents
Proposals, reports and meeting notes have to be read, but only a small part of them actually needs a decision.
What is included
- A look at what genuinely repeats in your working week today
- An honest read on which tasks suit this and which do not
- One task set up first, with the time it takes measured before and after
- Connected to the tools you already use, not a new system to learn
- A person approving before anything reaches a customer
- A review of where the data goes and what the provider does with it
- Documentation so you can switch it off again without calling anyone
How it works
We look at your week
Which tasks repeat, how often, and how long they actually take. Measured with a clock rather than estimated.
We pick the one that annoys you most
Not the one that sounds most impressive. The one that repeats often, has clear rules, and where a mistake gets caught rather than going straight to a customer.
We let it run for two weeks
It runs in parallel while you still check everything by hand. After two weeks you know whether it saves time, and where it gets things wrong.
We widen it or we stop
If the answer is no, we stop there. If it is yes, we take the next task. It is a cheap way to find out.
What drives the price
It starts with the mapping, meaning working out what the task actually consists of. From there the price follows how many systems have to talk to each other, whether they already have an interface to work against or everything has to go the long way round, and what the existing data looks like. Messy data is usually what takes the most time.
Then comes the part about making it safe. What security requirements apply to the information involved, how much testing it takes before we trust the result, and how badly it could go if something is wrong. The larger the consequence, the more checking gets built in, and the longer it takes.
Running costs come on top, because the AI service itself charges per use. That figure depends on how much goes through it, and you get an estimate with the proposal. The same goes for looking after it over time, since anything running every week needs somebody keeping an eye on it.
You get scope and price in writing before anything is set up.
Common questions about AI automation
What does AI automation mean in practice?
A task you do by hand today gets help from a tool. Drafted replies, sorting of incoming messages, summaries of long documents, tidying of submitted forms. A person always approves before anything reaches a customer.
Which tasks are a bad fit?
Anything that requires somebody to take responsibility. Binding quotes, replies to complaints, personnel matters and anything touching health or safety. The same goes for judgements where context matters more than the text, such as an unhappy customer asking for something other than what they wrote.
What happens to our data?
That is a decision to settle before the setup, not after. If you send customer data to a service, you need to know where it sits, what is stored and what gets used for further training. We go through it together and choose from there.
Do we have to change systems?
No. The point is connecting to what you already use. A new system to learn usually costs more than the task it was meant to solve.
What if it does not work?
Then we stop. That is why we start with one task and measure it for two weeks. Finding out early is cheaper than building a system around an assumption.
Keep reading
Tell us about the task that eats time
Describe how it works today and you get an honest answer on whether it is worth automating.