Services
AI development
Built into the CRM, helpdesk and accounting tools you already run.
Industries
Project stories
Each project story records the business need, the approach, and the outcome.
AI development
Four kinds of AI work we take on most: agents that run a multi-step task across your systems, assistants that answer from your own documents, extraction that turns emails and PDFs into records, and LLM features built into the screens your team already uses.
What we offer
Multi-step assistants that can look things up, call your systems, and complete bounded tasks.
Agents are useful when a task needs several steps and several systems: checking an order, updating a record, drafting a reply, escalating when unsure. We design the tool boundaries, permissions, and logging first, then build the agent to operate inside them.
Custom applications built on large language models, with your data, rules, and interface.
A chat window is rarely the product. We build LLM-backed applications with proper interfaces, prompt and version management, evaluation sets, and cost controls, whether the model comes from OpenAI, Anthropic, Google, or an open-weight option hosted in your own cloud.
Question answering over your own documents and databases, with sources cited.
Retrieval-augmented generation lets a model answer from your policies, manuals, tickets, and records instead of general training data. We handle ingestion, chunking, embeddings, access control, and evaluation so answers are grounded and traceable to a source.
Classification, extraction, and routing for text that arrives in volume.
Support tickets, reviews, contracts, and forms all carry structure that people currently extract by hand. We build NLP pipelines that classify, tag, extract entities, and route text so downstream systems and teams get clean input.
Also
These usually come up inside one of the offers above rather than as a project of their own.
Drafting, summarising, and content generation features inside the tools your team already uses.
Detection, classification, and document understanding from images and video.
Models trained on your data for scoring, classification, and recommendation.
Forecasts and early-warning signals built into the dashboards people already read.
Deployment, monitoring, and retraining so models keep working after launch.
How an AI engagement runs
The same four stages as any BizHues engagement, with two additions: a fit assessment before discovery ends, and an evaluation set that the feature is measured against before and after launch.
Working sessions to agree the users, the decisions the software supports, the systems it touches, and what the first release must do.
Data model, integration points, and clickable screens for the main flows before any build time is spent.
Short cycles with working software on a shared staging environment at the end of each one.
Release with monitoring in place, hand over the code and infrastructure, then use production feedback to decide what changes next.
Foundations
Reliable data, working integrations, and a system to put the feature into. These services usually run alongside AI work.
Questions
Describe the task, the data you have, and who would review the output. We tell you honestly whether a model is the right tool and what the first step would be.
Talk to an engineer
Book a 30-minute call with an engineer, at no charge. Tell us the problem and the systems involved, and you leave with questions answered and a suggested first step, not a sales deck.
