Michal Kaszubskikaszub.ski

Work · Products · 02 / 07

Owning the whole problem: an AI consultancy from discovery to deployment

NeuralTakeThe AI consultancy I founded in London in 2024, through which I have signed and delivered more than £5,000 of B2B AI software and consultancy work and shipped a production SaaS used by about 50 business users.

The problem

NeuralTake is the AI consultancy I founded in London in 2024; the domain was registered that September. Its current positioning is automation for professional services firms: workflow audits and fixed-fee builds. The work I want to show here is not one system but the combination that a small consultancy demands: the engineering and the client ownership, at the same time, by the same person.

Every engagement had the same shape. A business wanted AI in a workflow, and somebody had to own the whole problem: stakeholder discovery, scoping, requirements, architecture, delivery, onboarding, support and roadmap. That somebody was me.

What I did

  • Founded and led the company. Secured and delivered more than £5,000 in signed B2B AI software and consultancy contracts. That figure is signed value, not revenue, and it is stated that way everywhere on this site.
  • Owned the client relationship end to end. Discovery, scoping, requirements, architecture, delivery, onboarding, support and roadmap.
  • Designed and deployed a production AI SaaS. Built in Django, React and PostgreSQL, it supports AI-assisted generation, multilingual workflows, Stripe payments and seven third-party integrations (Facebook, Instagram, LinkedIn, X, TikTok, YouTube and Pinterest), plus WordPress blog publishing. It is used by approximately 50 business users and is publicly named AI Zuzi.
  • Took a client platform from prototype to production. A security audit, encryption at rest, multi-factor authentication, GDPR-aligned deletion and a zero-data-loss live migration from SQLite to PostgreSQL.
Feature panel of the AI Zuzi Socials Manager landing page: the headline 'AI Zuzi Socials Manager. Create. Schedule. Automate.', a short description, and four features named Compose, Schedule, Create and Measure, on a dark blue ground.
The AI Zuzi Socials Manager landing page, cropped to the feature panel: compose, schedule, create and measure. The sign-in form and a pilot user's testimonial have been cropped out.Screenshot, socials.aizuzi.com, September 2026

Decisions

  • A stack that is easy to reason about, deploy and support. Django, React and PostgreSQL. In a consultancy the person who builds the system also answers the phone, so operability counts as a feature.
  • Payments and integrations as first-class work. The SaaS earns its keep through Stripe and through the seven platform integrations, so those paths were designed and tested rather than bolted on.
  • Security and data protection before scale. The client platform went through a security audit, encryption at rest, MFA and GDPR-aligned deletion before it was called production.
  • Migrate live, and lose nothing. Moving a running system from SQLite to PostgreSQL without downtime or data loss was planned as its own project, with verification on both sides.
  • Scope in fixed fees and audits. The commercial shape on neuraltake.com, workflow audits and fixed-fee builds, keeps engagements bounded for the client and for me.
The neuraltake.com homepage: a dark hero reading 'Reclaim Manual Workflows' above a line about automating document-heavy workflows for professional services firms, and a button to book a 30-minute workflow audit.
The neuraltake.com homepage: automation for professional services firms, with a workflow audit as the first step. The cookie banner has been cropped away.Screenshot, neuraltake.com, September 2026

Evidence

Evidence, with context and dates
FigureContextAs of
£5,000+Signed B2B AI software and consultancy contracts, delivered; signed value, not revenue15 September 2026
About 50Business users of the production SaaS (AI Zuzi) I designed and deployed15 September 2026
7Third-party integrations in the SaaS: Facebook, Instagram, LinkedIn, X, TikTok, YouTube and Pinterest, plus WordPress blog publishing
0Records lost in the live SQLite-to-PostgreSQL migration of a client platform

The contracts are commercial validation: people paid for the work and it was delivered. The SaaS is a production deployment: it runs, takes payment and is used by about 50 business users. Neither figure is large, and neither is dressed up as more than it is.

Limitations and status

NeuralTake is a young, small company. The contract figure is signed value, not revenue, and the user figure is approximate. Client and pilot organisations are not named here, and no confidential screenshots are shown. Whether AI Zuzi is the “client platform” described above is being confirmed and is deliberately not stated.

What I learned

  • Discovery is engineering. Most of the risk in a build is decided before the first line of code, in what the client actually needs versus what they asked for.
  • Integrations are where the maintenance lives. Seven platforms means seven sets of changing APIs, and the support plan has to assume that.
  • A live migration is a product, with its own requirements, rehearsal and verification.
  • Being both the engineer and the owner is a constraint that improves the engineering: every shortcut comes back to you.