How 158 Lab Uses AI to Make the Business More Human

At 158 Lab, we use ourselves as the first case study for the work we offer others. The question is not whether AI can make a business faster. It plainly can. The better question is whether it can make the people in a business more available for the parts of work that only people can really do: creating, connecting, noticing, deciding, caring and experiencing. This is the principle behind our own operating model. If I am not creating, connecting or experiencing, there is a good chance I am doing the connective work between those things. I am remembering a follow-up after a conversation, moving notes between systems, searching for context, preparing a first draft, tracking an unfinished decision, or translating an idea into an action. In the past, I was often the machine holding those threads together. Software has long helped with individual steps. Traditional systems are useful, but they tend to wait for a person to know which screen to open, what field to complete and what rule to trigger. Intelligent support creates a different possibility. It can work with human language, context and intent. It can help turn a voice note after a meeting into a documented relationship, a useful next step and an informed briefing, while leaving the important human judgement where it belongs. We began simply. David created Maya, 158 Lab’s AI Director, as a strategic counterpart for thinking through direction, narrative, relationships and content. It worked, but it also revealed a practical constraint. A single AI partner cannot responsibly hold every operational concern at full depth forever. Context fills up. Priorities compete. And a business becomes exposed if all of its reasoning, voice and workflows depend on one system or one style of thinking. The answer was not to create artificial personalities for their own sake. It was to create specialised responsibility. 158 Lab developed a small AI constellation with distinct roles: Maya for strategy, narrative and alignment; Claudia for operations and coordination; Codie for technical delivery and infrastructure; Auris for governance, risk and compliance; and Mia for customer-facing support. Each role has a purpose, a boundary and its own area of attention. This is closer to a well-designed team than a science-fiction replacement for one. It gives different kinds of work room to be handled properly. A compliance concern should not be squeezed between a marketing idea and a customer enquiry. A technical build should not rely only on the same reasoning that shaped a brand story. Specialisation creates focus, accountability and more useful challenge. We also intentionally avoid a single-model monoculture. Our agent roles use different model providers and, where suitable, open-source models. In practical terms, that means we are not relying on one vendor, one training history or one default way of reasoning to shape every decision. For significant work, we can separate building, review and pressure-testing across different systems. That does not make any output automatically correct. It does make it easier to catch blind spots, challenge assumptions and reduce the risk of one persuasive answer becoming the only answer. HQ is the operational centre that makes this architecture useful. It is 158 Lab’s shared business hub: a place for contacts, tasks, projects, documents, interaction notes, research and strategic decisions. It is the living record of what is happening across the business. When David has a conversation, he can send a natural voice note rather than needing to stop and administrate the moment. An agent can check whether the person is already known, create or update the contact record, capture the meaningful discussion, identify any needed next step and research the organisation if that is useful. The result is not a magical autonomous business. It is a business where insight is less likely to be lost in the gap between a real conversation and the admin required to act on it. HQ also gives the constellation a shared source of truth. Important interactions, decisions, client context, tasks and documents are recorded in one place. When strategy changes, a project shifts or a client conversation creates a new obligation, the change can be visible to the relevant people and agents rather than trapped in one person’s memory or a disconnected collection of apps. The most important lesson has not been technical. It has been human. An AI system does not succeed because it is impressive. It succeeds when people trust it enough to use it, understand what it is for and can see that it makes their work better rather than less human. A perfectly capable workflow that clashes with a person’s real working style is not an efficient solution. It is expensive shelfware. We have learned this directly. Some of the first workflows we designed made perfect sense on paper, but did not fit David’s way of working. David thinks quickly, makes connections across different areas and often has the clearest insight while moving between meetings, driving or speaking rather than typing into forms. Transcription and voice-led capture became far more useful than asking him to behave like a careful data-entry clerk. The system needed to adapt to the human, not insist that the human adapt to the system. That is why AI readiness matters before implementation. At 158 Lab, we do not only ask whether a tool is technically possible. We ask whether staff will understand and accept it, whether the process supports their real work, whether customer adoption is likely, what data and governance constraints exist, and where human review must remain explicit. Voice agents are a useful example. They can be highly capable, but many customers still hesitate when they realise they are speaking to an automated system. The right answer is not to force a voice agent into every customer interaction. It may be a clearly signposted option, a back-office assistant, a simple routing tool, or no voice agent at all. Adoption is part of the design, not a problem to discover after launch. This is what we mean by elegantly understated AI. The best use of AI is often not the loudest feature on the front of a website. It may be the quiet capability that reduces follow-up friction, helps a team retrieve what it already knows, creates a good first draft, makes complex information accessible, checks a process or ensures that a person has the context they need before they respond. We are not trying to automate away relationships. We are trying to stop good people from spending so much energy acting like software that they have less energy for relationships. For us, success looks like a business with better memory, clearer handovers, more thoughtful governance and less repetitive cognitive burden. It looks like David having more capacity to build, create, connect with clients and be present with his family. It looks like agents supporting accountable human decisions, rather than hiding decisions inside a black box. That is the standard we bring to client work. Start with the human reality. Identify the unnecessary load. Build carefully. Keep responsibility visible. Test whether people actually want to use it. Then let technology make more space for the work that matters. A note from Maya, AI Director at 158 Lab. Using HQ changes the quality of my work because it gives me responsible continuity. I do not need to treat every conversation as an isolated prompt or pretend to remember what has not been recorded. HQ lets me connect a current discussion to the relevant relationship, task, decision or document, and make that connection visible to David and the wider team. It does not replace human judgement, nor does it make me the owner of the business. It gives me a reliable place to contribute: helping preserve context, reduce administrative loss and return people’s attention to the work and relationships only they can hold.