policies.

These policies help define how I work. I share them with my clients, although they're actually mostly for me.

I believe writing is thinking, and the act of writing these down - giving some shape to otherwise abstract thoughts about how I want to operate - was the useful part.

Feel free to use or adapt these as you see fit.

AI

June 2025

The current set of AI tools are quite remarkable and require thoughtful application, especially in a professional setting. As a “creative” whose work has almost certainly been used to train the current crop of Large Language Models, a former tech sector employee who helped launch the most popular open-source generative AI models (and a hobbyist who recreated and trained my own version of Karpathy’s nanoGPT on a custom corpus) I have a complicated & overlapping set of feelings about generative AI.

With any leading edge technology it is foolish to make predictions too far or firm. But there are some initial principles that are worth setting out regarding how I think about AI and my work, and which form the basis of this policy:

I will not provide or charge clients for work which is generated or edited by artificial intelligence, for the following reasons:

• My value lies in the quality of my thinking, the distinctiveness of my prose, the depth and breadth of the real world experience that informs my work, and the calibre of the advice I offer. I am an expert communication advisor and a world class writer: that is what you pay me for, and that is what you will get.
• It is wrong to charge people for what they could otherwise achieve with minimal effort from publicly available AI products. Generative AI is substantively different from other categories of tools, as it enables the creation of entire bodies of work based on only short and high-level instruction. Regardless of how sophisticated the prompting, training or deployment is, I am not interested of being a middleman between clients and AI models, and I believe it is unacceptable to pass models’ output off as your own work.

I use AI products for research, much the same way as I use search engines, articles, books and bibliographies: namely, as a means for finding relevant information and analysing existing research. Here AI acts as a kind of research assistant: fetching information based on my specific asks, which I then review and may use to inform my work, with one crucial caveat: until I feel confident that they have surpassed current accuracy standards for peer resources (i.e., their output is broadly as authoritative as Wikipedia etc.), I will continue to manually verify information provided by AI tools. This includes but is not limited to requiring citations and links, reviewing quotes or statements of fact against primary sources, and cross-checking against credible external resources.

I do not enter sensitive, confidential or potentially identifying client information into third-party AI models hosted on external servers or public clouds. However, I reserve the right to operate a private instance of a customised open-source LLM on my own secure server which I may use for work purposes, subject to the exclusions above; and to develop my own proprietary models or applications for research and analysis.

Conflicts

June 2025

I am free to decide who to work for and on what basis.

I have a responsibility not to take on clients or projects where

• there is the potential for meaningful commercial or competitive conflict (i.e., where clients will be in direct competition with one another, either on specific projects or at the organisational level);
• existing clients’ remit or interests are directly opposed (i.e., a regulatory body and a member of the industry they oversee); or
• there exists significant ideological tension (i.e. a potential client clearly stands on the other side of a major issue or argument from an existing one)

I retain the right to refuse or halt any individual project or client relationship if I conclude there is the potential for significant conflict.

In the event that a potential for conflict emerges, I have a duty to inform any affected clients and recuse myself if necessary. If in doubt, I will err on the side of caution.

When seeking to resolve conflicts, the primary principle is precedence: the client I signed a current contract with first gets first right of refusal, and can retain my services exclusively (potentially for an additional fee – subject to negotiation – to compensate for lost revenue), or terminate the relationship as per the contract.