Responsible AI and Technology

We use AI where it does something useful: accelerating research, drafting, summarising, structuring information, analysing patterns in data, and building systems that handle work that used to need a person for every instance.

We do not use it to replace judgment about a business. A diagnosis is a claim about where a company is losing value, and that claim needs a person who has looked at the business and is willing to be accountable for it.

AI-generated material is reviewed by a person before it is used for anything that matters: a diagnosis, a recommendation, a proposal, a piece of code going into production, a message sent under a client's name, or a number a decision will be based on.

Where we build automated systems for a client, we design a review point into them. The right question is not whether a system can run unattended, but what it should escalate and to whom.

The same tool is appropriate in one business and reckless in another. Regulated sectors, sensitive customer data, safety-related processes and contractual restrictions all change what is sensible.

We ask about those constraints before we design around them, and we would rather recommend a duller system that fits than an impressive one that does not.

We handle client information carefully and share it internally only with the people doing the work.

We do not knowingly submit confidential client information to public AI models for training, and we do not knowingly use one client's confidential information to build something for another. Where a tool's terms would allow content to be used for training, we configure it otherwise where that option exists, or we do not put confidential material into it.

We try to work with the least sensitive information that will do the job. In practice that means:

• Using samples, extracts and anonymised or aggregated data where they are enough
• Removing identifiers that the task does not need
• Keeping access to systems and accounts limited to what the work requires, and handing it back when the work is done
• Preferring tools where data handling is documented over tools where it is not
• Asking before introducing a new tool into a client environment

We use AI services operated by other companies, and we may change providers as the field moves. Each provider has its own terms, data practices and retention behaviour, and those terms govern what happens inside their systems.

We choose providers with that in mind, but we do not control them and we cannot give assurances on their behalf. Where a specific provider matters to an engagement, we will name it as part of that engagement.

AI systems produce outputs that can be inaccurate, incomplete, biased, outdated, or confidently wrong. They can misread context, invent detail, and reproduce patterns present in their training data.

This is not a reason to avoid them. It is a reason to build around them properly: constrained inputs, review points, monitoring after deployment, and a clear answer to what happens when the system is wrong.

Clients remain responsible for their own operational, legal, financial, regulatory and employment decisions. We can build a system, explain what it does and what it cannot do, and recommend a course of action. We cannot take on responsibility for decisions made inside your business.

Where a system we build will influence decisions about people, money or compliance, we will say so plainly and expect the business to apply its own oversight.

Where AI use is material to an engagement, we disclose it: what is automated, what is reviewed, what data is involved, and which categories of tool are used. If you would prefer a piece of work to be done without AI assistance, say so and we will tell you what that changes in cost and timeline.

This statement describes how Ayush Malik, working under the name North, approaches AI in client work. It is not a contract, and specific obligations for an engagement live in the agreement for that engagement.

• business@joinnorth.in

  Questions about how we use AI on a specific piece of work.