Professional services · MB Test Partners (Sydney)
Up to 40% faster test preparation: MB Test Partners and IntelliTest AI
A 30-person SAP testing consultancy turned decades of method into IntelliTest AI, a private AI platform on its own servers. Built as a fixed-price MVP in 20 business days, with working chat in the first week, and now in phase four.
The challenge
MB Test Partners is a Sydney software testing and QA consultancy of about 30 people, delivering SAP test verification for clients running large, complex SAP programs, backed by more than 30 years of testing experience. Like most professional services firms, its value sits in its people: their methods, test assets and judgement. • Expertise did not scale: quality depended on which consultant was on which site. • Preparation took time: drafting and reviewing test strategies and test cases took days of skilled effort on every program. • Trust mattered: in testing, an answer is only useful if you can check where it came from. • Confidentiality was non-negotiable: SAP programs run on client documents that cannot go into public AI tools. Data had to stay under MBTP's control and in Australia.
Our solution
MBTP wanted working software before committing to a larger investment, so we started with a fixed-price, fixed-scope MVP delivered in 20 business days. A working chat was running on MBTP's own documents within the first week. MBTP brought the testing method, domain knowledge and product vision; we designed and built the platform with them in short, reviewed sprints. • Quality you can check: consultants upload PDF, Word, Excel and PowerPoint files (including scans), ask questions in plain English and get answers that cite their sources. Always-on reference documents (policies, methods, style guides) keep every answer aligned with how MBTP works. • Expertise captured and shared: workspaces by team or client, with instructions, documents and access flowing down from parent workspaces, plus role-based access. • Control and governance: commercial frontier models or fully offline open-source models when data security is paramount; spreadsheet exports, an admin dashboard, and separate staging and production with an approved release process. • Teams of specialist AI agents: a researcher, analyst and reviewer can work as a pipeline, in parallel, by hand-off, or as a moderated discussion that produces an agreement score, so consultants know exactly what needs expert attention. In early 2026 the platform was rebuilt as version 2 for a stronger foundation, followed by a training guide for MBTP's team.
The results
• Up to 40% faster test preparation, with drafts grounded in MBTP's method and the client's own documents. • 80% to 90% accuracy on first-pass test case generation, needing only minimal refinement. • Private AI on MBTP's own servers in Australia, where public AI tools were previously off limits. • MBTP's clients commonly report saving up to 70% of time compared with manual methods, with tangible benefits in the first week. • Newer consultants produce senior-quality work sooner, because the firm's method is built into every answer. What's next: MBTP started with one fixed-price MVP and has kept building on what works. Phase four continues on the same foundations: private data, cited answers and specialist AI agents.