The approach

Small organisations can adopt AI faster. Decisions can be made more quickly, there are no compliance teams, and knowledge and use case sharing are easier in smaller organisations.

Technically, there are fewer systems to connect to, and the knowledge base is normally smaller, so it can be more easily used with AI to give context on the specific business activities and how it works.

Agents are popular right now. Many businesses want them. The reality of agents, and anything else built with AI that works autonomously to fulfil a function, is that they need maintaining. Models get retired and replaced, sometimes quickly. Tools change, and so does the data systems work with. Credentials expire.

For repeated, specific tasks, an automated workflow can beat an agent on reliability. Often the most reliable systems contain less AI, not more: plain software carries the routine, and the AI goes only where judgment is needed. Agents and AI systems have to be valuable in the work they are doing and worth maintaining. They should be well documented, so they are easier to fix, and secure, so they can be trusted. Part of this is hosting: where the system runs, on a cloud server or locally on a computer, because that decides what it costs, what happens when it breaks, and where your data sits.

With AI models' ability to write code, software is much easier to build, and it can become tempting to replace every software as a service (SaaS) product paid for by a business with something custom made. That is usually a mistake.

Building is tricky and time consuming, and every system built is another system that has to be maintained. A good product that already does the job well is normally better than rebuilding something that is already working. Custom software and agents do have a place: they fit where there are no existing solutions, and can bring real benefit where a business has specific needs. The skill is knowing what to build and what not to build.

It is an easy trap to fall into, and one Bigrise has been guilty of before.

A lot of AI work is done in connecting everything together securely, through MCPs and APIs, so that AI tools, agents and software platforms can access what they need. It is deciding what AI can reach, where its knowledge comes from, and how the data it relies on is organised.

AI works best when the digital side of the business is not complicated. Most businesses have messy data: hundreds of files and spreadsheets that are confusing and hard to read, formatted so that only a few people can understand them. Organising them, or getting AI to help organise them where it is safe to do so, is an important part of giving AI context on your business. Keeping key information in plain markdown files, a format AI reads easily, is one simple way to do this.

Businesses getting more from AI are keeping better information. AI does not need bigger models to understand most organisations' needs, it needs a map: what the business does, how it works, what it offers, how it is sold, your marketing channels and customers, written down somewhere it can read. Smaller organisations can capture key information in a small amount of files.

Every business needs a source of truth: one place where the facts live, kept current, that people and AI both trust. At Bigrise that is a single master.md file holding the map: where every other file lives and what is in it, with smaller files underneath for each part of the business. When the same fact appears in more than one place, every copy is kept matching. The AI tools Bigrise uses keep this up to date automatically: when something is added or removed, the map is updated too, so it stays true.

The biggest gains from AI come from using the tools well: knowing how they work, what they are good at, and where they fall short. That knowledge comes from use. Learning the tools first also makes every later decision better: what to buy, what to build, and what to leave alone.

The same applies when building. Building with AI does not always require a deep understanding of code, but it does require technical literacy and an understanding of how AI works. Building and deploying AI and software should be done cautiously: adequate security, an understanding of the data, and knowing what the systems are doing and how they run are all important for deploying something safely.

AI can be used for much more than another search engine or a way to draft emails. Used creatively, it can sit inside the work and provide business functions. Finding those use cases is the hard part, and often they are already known to the people doing the work, who do the task every day and have the insight and detail needed to automate it. The harder part is knowing the changing strengths and limitations that AI currently offers. That is the core of the Bigrise approach: understanding where the team's time is spent and the core functions of the business, then judging where AI can help, and where it cannot.

Not everything should be automated. Automations can be costly, and some perform worse than the manual process they replace. The right question is not what can be automated, but what is worth automating.

Keeping data safe is very important for all businesses. The question is: what happens to your information when it is put into AI tools? Paying for AI does not buy privacy. The consumer versions of the big tools, free or paid, can use the data put into them to train models, so it is not safe. Business accounts are usually different: most commit to not training on the data, but it is worth checking the right licences and the right settings. Knowing where the data is stored and what your AI tools have access to is very important. For UK businesses dealing with European businesses, the EU AI Act is a separate set of rules to be aware of.

The same thinking applies to access. What access do your AI tools and agents have, and what data goes into them? Even on paid business licences, this should be decided carefully. Agents can be manipulated into acting outside what they were intended to do, so what they can reach matters more than what they are told. You are responsible for the input into your AI tools and the output of the AI tools.

Shadow AI, AI being used in a business without being authorised, is a large part of the reality today. People putting data into free tools, or using unapproved AI at work, can potentially leak business data. Approved tools, set up safely, with a clear AI policy, can help navigate shadow use.