https://www.m3ca.com/assets/journal-CINmgjzx.js

1var e=[{slug:`training-ai-on-company-data`,title:`Training AI on company data: what to get right before you start`,date:`2026-10-08`,tag:`AI`,excerpt:`A practical guide to permissions, data preparation, retrieval, fine-tuning and evaluation before company information is connected to an AI system.`,author:`M3CA AI`,readingTime:`9 min read`,image:`scene-atlas-permissions`,body:[{type:`paragraph`,text:`When a team says it wants to train AI on company data, it can mean three very different things: letting a model retrieve approved documents at answer time, fine-tuning a model on examples of a specialist task, or training a new model. Those routes have different data, infrastructure and governance requirements. Treating them as one purchase is how promising pilots become expensive risks.`},{type:`heading`,text:`Begin with the task, not the archive`},{type:`paragraph`,text:`Do not begin by moving every file into a training folder. Define one bounded task, the person accountable for it, the input they may use and the output they need. Record how the task works today so the AI version has a baseline. If nobody can agree what a correct result looks like, more data will not solve the problem.`},{type:`heading`,text:`Decide what is allowed to leave the room`},{type:`paragraph`,text:`Classify the material before it reaches a model. Public, internal, confidential, personal and regulated records need different treatment. Check ownership, consent, retention, provider training terms, data location and the access rights of each user. Remove duplicates and obsolete policies before they become confident but outdated answers.`},{type:`list`,items:[`Who owns every dataset and example?`,`Which users may retrieve each document today?`,`Can a person request correction or deletion?`,`Where are prompts, outputs and logs retained?`,`What must never be sent to an external model?`,`Who approves the answer before it affects a customer, employee or regulated decision?`]},{type:`heading`,text:`Retrieval, fine-tuning and foundation training are not interchangeable`},{type:`paragraph`,text:`Retrieval is usually the first route for company knowledge because the source can be updated and cited without changing the model. Fine-tuning is useful for a repeatable specialist behaviour when you hold enough high-quality examples and can test on examples the model never saw. Training a foundation model requires a defensible dataset, substantial compute, research capacity and a reason existing models cannot meet the requirement.`},{type:`quote`,text:`The right question is not how much company data can be added. It is how little approved data is needed to prove one useful result.`},{type:`heading`,text:`Build the evaluation set before the model`},{type:`paragraph`,text:`Create representative questions, difficult edge cases, refusal cases and known correct answers before development begins. Keep some examples out of the build process. Score source support, factual correctness, missing information, review effort and safe refusal. Re-run the same set when the model, prompt, data or permissions change.`},{type:`heading`,text:`Choose the deployment from the risk`},{type:`paragraph`,text:`An approved business workspace may suit low-risk internal drafting. A controlled integration may suit company knowledge where the provider terms and access model are acceptable. Restricted workloads can justify self-hosting or an isolated environment. M3CA Sovereign AI provides a UK private-compute route for organisations that need their models and data away from the public internet.`},{type:`heading`,text:`Leave a system people can operate`},{type:`paragraph`,text:`The final deliverable is not only a model. It is an operating guide: owners, approved sources, user permissions, evaluation results, review steps, monitoring, incident handling, deletion and rollback. Train the people who will challenge the system, not only the people excited to use it.`},{type:`paragraph`,text:`M3CA helps teams choose between retrieval, fine-tuning and private deployment, then builds and evaluates the narrowest useful workflow. Bring one task and a safe sample to an AI consultancy session in our Mayfair studio.`}]},{slug:`ai-consultancy-london-after-the-contract`,title:`AI consultancy in London: what should happen after the contract is signed`,date:`2026-10-08`,tag:`AI`,excerpt:`The practical work an AI consultancy should own after the pitch: baselines, data boundaries, evaluation, integration, staff enablement and handover.`,author:`M3CA AI`,readingTime:`8 min read`,image:`scene-consulting`,body:[{type:`paragraph`,text:`An AI consultancy is easy to judge during the pitch and harder to judge once the work starts. The useful test is not how many models appear in the deck. It is whether the consultancy turns one important workflow into a system your people can inspect, operate and improve without depending on a permanent demonstration.`},{type:`heading`,text:`Week one should produce a baseline`},{type:`paragraph`,text:`The team should observe the task as it works now, including exceptions and handovers. Agree representative examples, elapsed time, quality checks and the cost of an error. Without that baseline, every later claim of improvement is only an impression.`},{type:`heading`,text:`The data boundary should be written down`},{type:`paragraph`,text:`You should know which sources are approved, where prompts and outputs are retained, whether provider terms permit the data, how user permissions are enforced and what the system must refuse. A London meeting is useful for getting the legal, operational and technical owners around one table. The postcode itself proves nothing.`},{type:`heading`,text:`A pilot should fail in public`},{type:`paragraph`,text:`A credible pilot includes missing documents, conflicting sources, unauthorised users and questions outside scope. Ask the consultancy to show those failures and explain the stop condition. The narrow system that refuses safely is often more valuable than the broad assistant that answers everything.`},{type:`list`,items:[`A named senior person remains accountable after discovery.`,`Model choice follows testing on your tasks rather than a provider preference.`,`Acceptance criteria are agreed before the demonstration.`,`Integration and permission work is visible in the plan.`,`Running cost is tested at realistic usage.`,`Staff training and operational ownership are part of handover.`,`The consultancy is prepared to recommend buying an existing tool or stopping the project.`]},{type:`quote`,text:`Consulting earns its place when the recommendation survives the workflow, the permissions and the first difficult example.`},{type:`heading`,text:`Integration is where the value or the risk appears`},{type:`paragraph`,text:`Useful AI rarely lives in a blank chat window. It belongs beside the documents, CRM records, approvals and people already doing the work. That means identity, access, logging, human review and fallback are product requirements, not technical polish for later.`},{type:`heading`,text:`Handover should begin before launch`},{type:`paragraph`,text:`Your team should receive the prompts, evaluation set, system access, operating guide and named responsibilities. They should know how to review a poor answer, update a source, watch usage and reverse a change. Continued support can be valuable, but dependence should never be the business model.`},{type:`heading`,text:`What M3CA owns`}

1,{type:`paragraph`,text:`M3CA keeps strategy, engineering, integration and staff enablement together. We work across leading commercial and open models, use our own systems as a proving ground, and offer private UK infrastructure where the data requirement justifies it. Client evidence is published with names withheld until written permission is granted.`},{type:`paragraph`,text:`If you are comparing AI consultancies in London, bring one workflow to our Mayfair studio. We will help you decide whether it needs a tool, an integration, a specialist model or no AI at all.`}]},{slug:`software-studio-mayfair-what-you-are-paying-for`,title:`A software studio in Mayfair: what you are actually paying for`,date:`2026-09-19`,tag:`IT`,excerpt:`Why a Mayfair address matters less than who sits in the boardroom when your roadmap is decided, and how to test that in a single consulting session.`,author:`M3CA Engineering`,readingTime:`6 min read`,image:`scene-consulting`,body:[{type:`paragraph`,text:`We run our software studio from 35 Berkeley Square, and we are relaxed about the fact that a W1 postcode builds nothing on its own. What a Mayfair studio should buy you is proximity: senior people you can sit with on a Tuesday afternoon, in the same room as the decision, rather than a relationship conducted entirely through a project manager and a status report.`},{type:`heading`,text:`The boardroom test`},{type:`paragraph`,text:`Before you commission anything, ask for one consulting session where the roadmap goes up on the screen. Watch who speaks. If the people describing the architecture are not the people who will write it, you are buying a sales function and inheriting a delivery team you have never met.`},{type:`list`,items:[`Who, by name, is on this from week one to handover?`,`What decision are we making in this session, and what changes if we are wrong?`,`Which part of the roadmap is genuinely uncertain, and how do we buy that certainty cheaply?`,`What would make you tell us not to build this?`]},{type:`quote`,text:`A studio worth its address will talk you out of the wrong project before it quotes the right one.`},{type:`heading`,text:`What proximity is worth in delivery`},{type:`paragraph`,text:`On a clinic network rollout we sat in eleven receptions and watched the real working day before touching the design. That is the practical advantage of a central London base: the cost of turning up is low enough that you turn up. Remote-first delivery can be excellent, but it rarely catches the local habit that quietly breaks a rollout.`},{type:`heading`,text:`What it is not worth`},{type:`paragraph`,text:`It is not worth a premium on rates for the sake of the square. We quote every engagement from the brief, against named deliverables, and the address does not appear in the arithmetic. If a studio cannot separate its overhead story from your scope, ask for the scope again.`},{type:`paragraph`,text:`If you are shortlisting a software studio in Mayfair or anywhere else in London, bring the brief to a twenty-minute fit call. We will tell you plainly whether we are the right team for it.`}]},{slug:`choosing-a-digital-product-studio-in-london`,title:`Choosing a digital product studio in London: how to read a pitch`,date:`2026-09-19`,tag:`IT`,excerpt:`Nine things to look for when a digital product studio shows you user flows and wireframes, and the three answers that tell you whether a first release will ship.`,author:`M3CA Product`,readingTime:`7 min read`,image:`scene-ux`,body:[{type:`paragraph`,text:`London has no shortage of studios who can produce a beautiful wall of wireframes. Fewer can tell you which of those frames earns money, which is decoration, and which will be deleted after the first fortnight of real use. That distinction is the whole job.`},{type:`heading`,text:`1. Ask what the first release proves`},{type:`paragraph`,text:`A first release is an experiment with a bill attached. It should prove one thing: that people complete the journey, that the integration holds, that the operation can run it. If nobody in the pitch can state the proposition in a sentence, the scope is a wish list.`},{type:`heading`,text:`2. Watch how they handle the user flows`},{type:`paragraph`,text:`Good product designers argue with you about the flow and agree quickly about the visual. When we sketch user flows for a booking or intake journey, the interesting frames are the failure paths: the person who abandons, the record that arrives incomplete, the staff member cleaning it up on Monday.`},{type:`list`,items:[`Does the pitch show the unhappy path, or only the demo path?`,`Who owns the content that fills these screens, and does it exist yet?`,`What happens to the work already in flight on launch day?`,`How will we know in month two whether this worked?`]},{type:`quote`,text:`Any studio can show you the journey that goes right. Ask to see the one that goes wrong.`},{type:`heading`,text:`3. Insist on a baseline before you start`},{type:`paragraph`,text:`If there is no baseline, there is no result, only a story told afterwards. We agree the measurement period, the conversion events and the attribution method before a line of code is written, and we record the evidence behind any figure we later publish. It is slower to set up and much harder to argue with.`},{type:`heading`,text:`A sensible way to start`},{type:`paragraph`,text:`Start on one journey rather than a platform. One page, one story, one follow-up, measured against a baseline, is a far better test of a studio than a six-month programme you cannot cancel.`}]}

1,{slug:`what-a-first-release-should-cost-you-in-attention`,title:`What a first release should cost you in attention, not just budget`,date:`2026-09-18`,tag:`IT`,excerpt:`The client-side effort a London digital product studio should ask for up front: decisions, content, responsive review time, and who signs off on layouts.`,author:`M3CA Product`,readingTime:`5 min read`,image:`scene-web-design`,body:[{type:`paragraph`,text:`Most delayed projects we inherit were not short of engineering. They were short of decisions. A studio that only quotes its own effort is hiding the other half of the cost: yours.`},{type:`heading`,text:`The four things we ask for`},{type:`list`,items:[`A named decision maker who can approve responsive layouts without a committee.`,`Real content, or an honest admission that we are writing it, priced as work.`,`Two review slots a week, short ones, while the desktop and mobile frames take shape.`,`Access to the systems the new journey has to speak to, arranged before week one.`]},{type:`quote`,text:`A fortnight of waiting for sign-off costs more than the design it is holding up.`},{type:`heading`,text:`Why responsive review needs a human`},{type:`paragraph`,text:`We build layouts across desktop and mobile frames at the same time, because most of the traffic that matters arrives on a handset in poor light. Somebody on your side needs to look at both on their own device and say whether it reads. Screenshots in a document are not that.`},{type:`heading`,text:`What we do with the time you give us`},{type:`paragraph`,text:`We spend it narrowing. Every review either removes a frame or confirms one. By the end of a sprint the page is smaller than the first sketch, which is almost always the point.`}]},{slug:`custom-software-development-london-what-it-costs-you`,title:`Custom software development in London: what it really costs you`,date:`2026-09-08`,tag:`IT`,excerpt:`The three costs London buyers underestimate when they commission custom software, drawn from builds we have run across clinics, universities and professional services firms.`,author:`M3CA Engineering`,readingTime:`7 min read`,image:`scene-software-build`,body:[{type:`paragraph`,text:`Most briefs we receive in London are priced on build effort alone. The build is rarely what hurts. The three costs that decide whether a custom system pays for itself arrive after the first release, and they are all avoidable if you plan for them in week one.`},{type:`heading`,text:`1. The cost of the systems you already run`},{type:`paragraph`,text:`Custom software almost never lands in an empty room. It has to speak to a practice management system, a finance ledger, a rota, an identity provider. On a patient pathway we rolled across eleven clinic sites, the integration and cutover work was a larger share of the programme than the application itself, because every site had a slightly different local habit. Ask any London studio to price integration and migration as named lines, not as a percentage buried in a total.`},{type:`heading`,text:`2. The cost of the people who have to use it`},{type:`paragraph`,text:`A system that a team quietly works around has cost you twice. When we rebuilt a university certification track, the completion rate moved because the sequence matched how learners actually studied, not because the technology was newer. Budget for research and for the second pass after real use. It is cheaper than a rebuild and far cheaper than a system nobody adopts.`},{type:`heading`,text:`3. The cost of proving it works`},{type:`list`,items:[`Evidence: audit trails and exports you can hand to a regulator or an auditor without a fortnight of screenshots.`,`Operations: backups you have actually restored from, alerting on the transactions that earn money, patching on a schedule.`,`Ownership: your code in your repository, your data in your account, documented well enough that another team could take it on.`]},{type:`quote`,text:`If the contract does not say who owns the code and the data, that is the most expensive line in it.`},{type:`heading`,text:`When custom is the wrong answer`},{type:`paragraph`,text:`We tell London clients to buy off the shelf whenever the process is genuinely standard. Custom development earns its keep when the process is your competitive advantage, when compliance forces a shape no vend

1or supports, or when you are paying per seat for software that fits nobody. If none of those is true, we will say so in the first conversation.`},{type:`heading`,text:`What to ask before you sign`},{type:`list`,items:[`Who, by name, writes the code, and will they still be on it in month six?`,`What is in the first release, and what does it prove?`,`How is integration and data migration priced separately?`,`What happens on day one after handover: who deploys, who is on call, who holds the keys?`]},{type:`paragraph`,text:`We run custom software development from the Mayfair studio and work on site with teams across London. If you have a brief, send it and we will tell you honestly whether it needs building.`}]},{slug:`choosing-an-ai-agency-in-london`,title:`Choosing an AI agency in London: eleven questions that separate the real ones`,date:`2026-08-19`,tag:`AI`,excerpt:`How to interrogate an AI agency before you commission anything, including the evaluation, permission and cost questions most pitches avoid.`,author:`M3CA AI`,readingTime:`8 min read`,image:`scene-ai-retrieval`,body:[{type:`paragraph`,text:`Every agency in London now offers AI. Very few can show you what happens when the model is wrong, and that is the only part that matters once a system is live in front of staff or customers.`},{type:`heading`,text:`Ask about evaluation before you ask about models`},{type:`paragraph`,text:`An assistant is only as trustworthy as the test set behind it. On our own retrieval work we hold a fixed set of real questions with known correct answers, and every change is scored against it before release. If a prospective partner cannot describe how they measure quality, they are shipping vibes. Ask to see an evaluation run, not a demo.`},{type:`heading`,text:`Ask who is allowed to see what`},{type:`paragraph`,text:`Retrieval systems fail on permissions long before they fail on language. If a document is restricted in your file store, the assistant must respect that restriction for every user, every time. Permission modelling is unglamorous engineering work and it is where most proofs of concept quietly break when they meet a real organisation.`},{type:`heading`,text:`The eleven questions`},{type:`list`,items:[`What does your evaluation set look like, and who wrote the correct answers?`,`How do you handle a question the system should refuse to answer?`,`Where does our data go, and is it used for training anywhere?`,`How are document permissions enforced per user?`,`What is the fallback when the model is unavailable?`,`How do you log what the system said, so we can review it later?`,`What does this cost per month at our actual volume, not at demo volume?`,`Who owns the prompts, the pipeline and the index?`,`How will you tell us this is the wrong use of AI?`,`What happens after launch: who monitors quality drift?`,`Can we speak to a client running this in production?`]},{type:`quote`,text:`The useful answer to most AI briefs is a narrower system than the one being asked for.`},{type:`heading`,text:`Where AI has genuinely paid for itself`},{type:`paragraph`,text:`The wins we see in London are dull and specific: finding the right clause in a decade of documents, triaging inbound enquiries so a person answers the right ones first, turning an audit trail into an evidence pack, drafting from a house style rather than from nothing. Broad assistants impress in a boardroom and disappoint in a workflow.`},{type:`paragraph`,text:`We build AI systems from the Mayfair studio, alongside the engineering practice that has to run them afterwards. If you are shortlisting an AI agency in London, put these questions to all of us.`}]},{slug:`what-a-digital-product-studio-actually-does`,title:`What a digital product studio actually does`,date:`2026-07-22`,tag:`IT`,excerpt:`The difference between an agency, a dev shop and a product studio, and how to tell which one your brief needs.`,author:`M3CA`,readingTime:`6 min read`,image:`scene-client-portal`,body:[{type:`paragraph`,text:`Three kinds of company will answer your brief in London, and they behave very differently once the contract is signed.`},{type:`list`,items:[`An agency sells campaigns and pages. Judged on how it looks and how it launches.`,`A development shop sells hands. Judged on tickets closed against a specification you wrote.`,`A product studio sells outcomes. Judged on whether the thing works in the hands of the people it was built for.`]}

1,{type:`heading`,text:`Signs you need a studio, not a specification`},{type:`paragraph`,text:`If you can hand over a complete specification and be confident it is right, hire hands and manage them well. If the honest answer is that you know the problem but not the solution, a fixed specification will lock in the first guess. A studio expects the shape to change between discovery and release, and prices that reality rather than pretending it away.`},{type:`heading`,text:`What good looks like week to week`},{type:`list`,items:[`Something working in front of real users inside the first month, however small.`,`One senior person accountable, in the room, for the whole engagement.`,`Decisions written down where you can find them in a year.`,`A release you could stop after and still be better off than before.`]},{type:`quote`,text:`The test of a studio is whether the first release could stand alone if the budget stopped.`},{type:`paragraph`,text:`We run as a digital product studio in Mayfair: strategy, design, engineering and the film work in one building, with the same team through to launch.`}]},{slug:`moving-a-london-team-off-legacy-software`,title:`Moving a London team off legacy software without losing a day`,date:`2026-06-10`,tag:`IT`,excerpt:`Cutover planning, data migration and the rehearsal discipline that keeps operations running while the old system is retired.`,author:`M3CA Engineering`,readingTime:`6 min read`,image:`scene-data-analysis`,body:[{type:`paragraph`,text:`Replacing a system people depend on daily is a logistics problem wearing a software costume. The engineering is usually the easy half.`},{type:`heading`,text:`Rehearse the cutover, more than once`},{type:`paragraph`,text:`We moved four hundred terabytes for a production company across six weekends without losing a shooting day, and the reason was rehearsal, not bandwidth. Every migration got a dry run against real data, a timed rollback and a named person who could call it off. Nothing about that is clever. It is simply planned.`},{type:`heading`,text:`The rules we hold to`},{type:`list`,items:[`Run both systems in parallel until the new one has survived a full business cycle.`,`Migrate in slices the business understands: one site, one department, one service line.`,`Agree the rollback before the go-live, in writing, with a time limit.`,`Reconcile counts and money after every slice, not at the end.`,`Keep the old system readable for as long as your retention rules require.`]},{type:`quote`,text:`A migration without a rehearsed rollback is not a plan, it is a hope with a date on it.`},{type:`paragraph`,text:`If you are carrying a system nobody wants to touch, the first piece of work is not a rebuild. It is a map: what depends on it, what data has to survive, and what a safe first slice looks like.`}]},{slug:`what-a-production-day-costs`,title:`What a production day actually costs`,date:`2026-01-14`,tag:`Media`,excerpt:`Crew, kit, insurance, post and contingency, broken down honestly so you can read a quote properly before you sign it.`,author:`M3CA Production`,readingTime:`6 min read`,image:`editor`,body:[{type:`paragraph`,text:`Most production quotes hide the same five costs behind one number. Here is the breakdown we use internally, in the order it lands on a budget.`},{type:`heading`,text:`The five lines that move`},{type:`list`,items:[`Crew: the day rate is the smallest part; prep, travel and turnaround are the rest.`,`Kit: camera and lenses are predictable, lighting and grip are where a day quietly doubles.`,`Insurance and permits: cheap to buy, expensive to discover on the morning of a shoot.`,`Post: assume edit time is at least three times shoot time for anything narrative.`,`Contingency: ten percent, held back, spent openly or returned.`]},{type:`quote`,text:`A quote you can read is a quote you can argue with, and that is the point.`},{type:`paragraph`,text:`If a supplier will not itemise those five lines, the risk has not gone away. It has just moved to you.`}]},{slug:`your-website-is-an-it-system`,title:`Your website is an IT system, not a brochure`,date:`2025-12-02`,tag:`IT`,excerpt:`Uptime, backups, patching and access control matter more to your pipeline than another hero animation.`,author:`M3CA Engineering`,readingTime:`5 min read`,image:`office-meeting`,body:[{type:`paragraph`,text:`The site is usually the only system every customer touches, and the only one nobody has assigned an owner.`}

1,{type:`heading`,text:`Treat it like production infrastructure`},{type:`list`,items:[`Know who can deploy, and remove everyone else.`,`Back up content and media separately from code.`,`Patch on a schedule rather than after an incident.`,`Alert on the checkout or the contact form, not just the homepage.`]},{type:`paragraph`,text:`None of that is glamorous, and all of it is cheaper than the week you spend rebuilding trust after an outage.`}]},{slug:`podcasts-that-survive-episode-ten`,title:`Podcasts that survive episode ten`,date:`2025-11-08`,tag:`Media`,excerpt:`Format discipline, a repeatable studio setup and a clip pipeline are what keep a show alive past the launch spike.`,author:`M3CA`,readingTime:`4 min read`,image:`film-set`,body:[{type:`paragraph`,text:`Almost every show that dies does so between episodes six and twelve, when novelty runs out and process has not replaced it.`},{type:`list`,items:[`Fix the format before the first record, then leave it alone for ten episodes.`,`Keep the studio built. A setup that takes an hour will not happen weekly.`,`Cut clips in the same session as the episode, never later.`]},{type:`paragraph`,text:`Consistency is the whole product. Audiences forgive an ordinary episode and never forgive a missing one.`}]},{slug:`cyber-essentials-without-the-theatre`,title:`Cyber essentials without the theatre`,date:`2025-10-21`,tag:`IT`,excerpt:`The controls that genuinely reduce risk for a 20 to 200 person business, and the ones vendors sell you anyway.`,author:`M3CA Security`,readingTime:`7 min read`,image:`office-meeting`,body:[{type:`paragraph`,text:`Small organisations are not attacked by novel techniques. They are attacked by the same four routes, every time.`},{type:`list`,items:[`Phishing into an account without multi-factor authentication.`,`An unpatched device someone took home a year ago.`,`A shared admin login nobody has rotated.`,`A backup that has never been restored from.`]},{type:`quote`,text:`Close those four and you have removed more risk than any dashboard you can buy.`}]}],t=t=>e.find(e=>e.slug===t);export{t as n,e as t};