I help companies identify where AI and automation can save time, make knowledge easier to access or improve an existing process. That can be something relatively simple, such as automating a repetitive workflow, or a custom system that combines company data, AI models and existing tools.
Start with the work
There is no shortage of AI tools. The harder question is which of them are actually useful for your company.
I usually start by looking at how work is done today. Where does information get copied from one place to another? Which tasks keep coming back? Where do people spend time searching for information, preparing documents, answering the same questions or maintaining data manually?
These are often much better starting points than asking what you could do with the latest AI model.
Once we understand the process, we can decide what should be automated, where AI can help and which parts should stay exactly as they are.
Automating repetitive work
A surprising amount of work still happens between systems.
Information arrives through an email, someone copies it into another tool, creates a task, prepares a document, updates a spreadsheet and lets somebody else know that it happened.
Many of these processes can be simplified.
I work with the experts at PAKD to connect existing systems, APIs and AI models and turn manual workflows into software that runs reliably in the background.
That might involve processing incoming documents, preparing reports, enriching CRM data, creating tasks from meetings, generating drafts, categorising requests, synchronising information between systems or taking care of recurring administrative work.
AI does not have to be involved in every step. Quite often, traditional automation is more predictable and cheaper. We use AI where understanding language, interpreting information or making a flexible decision actually adds something.
Making company knowledge usable
One of the most useful applications of AI is giving people better access to knowledge that already exists inside a company.
That knowledge is often spread across wikis, project documentation, PDFs, shared drives, emails, support systems and people's heads.
Using Retrieval-Augmented Generation, usually shortened to RAG, we can connect language models to these sources and let people ask questions in natural language.
Instead of relying only on what a model learned during training, the system retrieves relevant information from your own documents and uses it as context for the answer.
This can become an internal knowledge assistant, a better way to search documentation, support for onboarding new employees or part of a larger application.
The interesting work is rarely the chat interface. Good results depend on the quality and structure of the underlying information, permissions, retrieval, source references and testing whether the answers are actually reliable.
Assistants and agents
Some systems can go a step further than answering questions.
An AI assistant might prepare something for a person to review. An agent can also interact with other systems, use tools and carry out several steps of a process.
For example, a system could read the notes from a meeting, compare them with information from a project, prepare follow-up tasks, draft an email and update another tool.
How much autonomy makes sense depends on the task.
For low-risk work, a process might run automatically. For anything involving important decisions, money, customers or sensitive information, it often makes more sense to have a person approve the result before anything happens.
We decide that as part of the system rather than after it has already been built.
Pilot first
I prefer starting with one concrete use case.
We build a small working version, run it with real data and compare it with the way the task was handled before.
Does it actually save time? Is the output good enough? How often does someone need to correct it? Does it fit into the existing workflow? Does the team want to use it?
If it works, we can expand it. If it needs adjustment, we learn that early. And if the idea does not deliver enough value, we know before it turns into a large AI project looking for a reason to exist.
From idea to working system
Depending on the project, my role can start with identifying possible applications, reviewing an existing idea or helping define the technical approach.
Together with PAKD, I can also take it through implementation. That includes prototypes, custom interfaces, integrations, RAG systems, knowledge bases, workflow automation, AI agents and the infrastructure around them.
I stay involved on the consulting side while bringing in the right specialists when deeper development or design expertise is needed.
The result should be something that makes everyday work noticeably easier and is reliable enough that people actually want to use it.