Frequently Asked Questions

About Brainwashed

What does Brainwashed do?

Brainwashed builds technology that solves real business problems. Our work falls into three areas: bespoke systems built from scratch around your workflows, custom integrations that connect your existing platforms into a unified ecosystem, and AI solutions that automate decisions, surface insights, and drive revenue. We have been established since 2013 and bring over 20 years of collective expertise in web design, development, and AI.

Who is Brainwashed for?

We work with businesses of all sizes — from growing e-commerce brands to mid-market and enterprise organisations — who need custom technology rather than off-the-shelf software. If your workflows are complex, your systems are disconnected, or you want to harness AI without compromising data security, we are the right partner.

How does Brainwashed approach a project?

We use Scrum and Agile methodology to deliver projects iteratively and transparently. You see progress every sprint and provide feedback early and often. We start by understanding your workflows, pain points, and data landscape, then architect solutions that integrate with your existing systems — no rip-and-replace, no disruption.

AI & LLM Training

Why would I train a language model on my own data instead of just using ChatGPT or Claude?

Generic models are trained on the public internet — they know a little about everything but not enough about your business. Fine-tuning on your own data gives you a model that understands your domain terminology, reflects your brand voice, hallucinates far less on topics you care about, and runs at a fraction of the cost of frontier models. Businesses typically see a 30–50% reduction in incorrect outputs after fine-tuning on domain-specific data.

What is the difference between a base model, an instruct model, and a thinking model?

A base model completes text without following instructions. An instruct model has been trained with RLHF to follow directions and hold conversations — this is what most production applications use. A thinking model (such as DeepSeek R1 or Claude with extended thinking) reasons step-by-step before answering, giving higher accuracy on complex tasks at the cost of more latency and cost.

What data can I train a model on?

Virtually any text-based data your business generates: customer support transcripts, product descriptions, internal knowledge bases, marketing emails, legal contracts, sales call notes, engineering documentation, and customer reviews. Quality matters more than quantity — a well-curated dataset of 1,000 high-quality examples can outperform 100,000 noisy ones.

Should I use local (self-hosted) AI or cloud AI?

For most structured, repetitive business tasks — support automation, summarisation, classification, and content generation — a self-hosted small or medium model delivers comparable results to frontier cloud models at 10–100x lower cost, with no data leaving your infrastructure. Cloud AI is still the right choice for burst workloads, experimentation, or tasks that genuinely require frontier reasoning. Many businesses adopt a hybrid approach: local for high-volume or sensitive workloads, cloud for complex edge cases.

Privacy & Data Security

What are the risks of sending my data to cloud AI providers?

Key risks include: data being transmitted to third-party servers and processed under their policies; the possibility of your inputs being used to train their models; regulatory exposure under GDPR, HIPAA, or PCI DSS when personal or financial data is involved; vendor lock-in when you build deeply on proprietary APIs; and an expanded attack surface from prompt-injection vulnerabilities. For sensitive workloads, self-hosted AI eliminates all of these concerns.

How does self-hosted AI help with SOC 2 compliance?

Self-hosted AI removes the need to add a third-party AI provider as a sub-processor in your SOC 2 scope. There are no third-party data flows to document, no vendor attestations to rely on, and no additional security perimeters to evaluate. You own the full audit trail — every inference, data access, and configuration change is logged in your systems. Organisations switching to self-hosted AI for sensitive workloads consistently report cleaner, faster SOC 2 audits.

MCP Servers

What is an MCP server and why does it matter?

The Model Context Protocol (MCP) is an open standard that lets AI models securely connect to your databases, APIs, and services. An MCP server wraps one capability — a database, a CRM, an e-commerce platform — and exposes it to any compatible AI client through a standardised interface. Without MCP, AI is a brain with no hands. With MCP servers, your AI can check stock levels, process refunds, update CRM records, and hundreds of other tasks without human intervention.

What systems can MCP servers connect to?

Virtually anything with an API or database: CRMs (Salesforce, HubSpot), ERPs (SAP, NetSuite, Odoo), e-commerce platforms (Shopify, WooCommerce, Magento), databases (PostgreSQL, MySQL, MongoDB), communication tools (Slack, Teams, email), file systems (S3, Google Drive), finance tools (Stripe, Xero, QuickBooks), and custom internal systems. If it has an API, a database, or a data feed, it can be wrapped in an MCP server.

Custom Integrations

Why do I need custom integrations when off-the-shelf connectors exist?

Off-the-shelf plugins handle basic connections but break down when your business has custom workflows, legacy systems, multiple sales channels, or specific data transformation requirements. Generic connectors sync the wrong fields, miss edge cases, fail silently, or cannot handle your volume. Custom integrations are built for your exact data model and edge cases. Every integration we build includes error handling, retry logic, logging, monitoring, and documentation.

Which CRM, ERP, and payment systems do you integrate with?

For CRMs: Salesforce, HubSpot, Microsoft Dynamics 365, Zoho CRM, and Pipedrive. For ERPs: SAP (S/4HANA, Business One, ECC), Oracle NetSuite, Microsoft Dynamics 365 Business Central, Odoo, Xero, and QuickBooks. For payments: Stripe, PayPal, Adyen, Worldpay, Klarna, Afterpay, and regional gateways including M-Pesa, PIX, UPI, and iDEAL. For fulfilment: ShipBob, ShipStation, DHL, FedEx, UPS, Royal Mail, and Amazon FBA. If it has a well-documented API, we can integrate it.

E-Commerce Development

Which e-commerce platforms do you build on?

We build on Shopify (including Shopify Plus and headless with Hydrogen), Magento 2 / Adobe Commerce, WooCommerce, Instagram Shops, and WhatsApp Business. Shopify is ideal for speed and managed infrastructure; Magento 2 for enterprise-grade flexibility and complex catalogues; WooCommerce for content-rich, SEO-driven stores; and Instagram and WhatsApp for social commerce where your customers already spend their time.

Can you migrate my store from one platform to another?

Yes. We handle migrations between any of the major platforms — Magento 1 to Magento 2, WooCommerce to Shopify, BigCommerce to Magento, and others — with zero data loss. This includes product data, customer records, order history, SEO URLs, and integrations. Migrations are planned carefully to minimise downtime and preserve your search rankings.

Bespoke Systems & Products

When does it make sense to build a bespoke system rather than buying software?

Bespoke makes sense when your workflows are complex and unique, you are running multiple disconnected systems that no single off-the-shelf product covers, existing solutions require you to compromise core processes, or you need full control over your data, logic, and roadmap. We handle the entire journey from defining requirements and writing the technical specification through to development, project management, and deployment.

What is Middle-man?

Middle-man is our SaaS platform built to solve one of the most painful problems in e-commerce: keeping all your systems in sync. Running an online business means juggling a storefront, a CRM, an ERP, and accounting software — and without a reliable bridge between them, your team wastes hours on manual CSV exports, data entry, and fixing sync errors.

Middle-man connects your platforms — Shopify, Magento, WooCommerce, and any CRM or ERP you rely on — so that orders, products, customers, inventory levels, pricing, shipments, and payment status all flow automatically in both directions. Every sync is version-controlled through Forgejo, giving you a complete audit trail of every change. If something goes wrong, you can roll back to any previous state instantly. Configure sync rules once, and Forgejo workers handle the rest on your schedule or in real time.

What is Backchat?

Backchat is our incentivised market research platform built to fix the biggest problem with traditional surveys: low response rates, bot submissions, and disengaged participants that produce data you cannot trust.

Backchat motivates real people to respond thoughtfully by allocating a portion of every response fee toward a prize pool — participants have a genuine reason to engage carefully. On the validation side, automatic bot detection filters out fraudulent submissions before they contaminate your data, and built-in control questions verify that each participant is paying attention. Invalid responses are flagged and excluded automatically.

Pricing is simple and transparent: a R5,000 setup fee covers your first 1,000 responses, platform setup, and prize pool allocation. Need more? Top up at any time — no subscriptions, no hidden fees. The result is clean, verified data from people who actually paid attention, ready for analysis.