Before you commit a budget to AI, you need to know what to build, what it will cost to run and whether it will do the job. Sphere 10 designs the whole solution around one useful business result, such as an answered enquiry, a qualified lead, a confirmed booking or an approved quote, and proves the uncertain parts in working software before you fund the full build.

What we do

Map the work. We trace the job through the people, records and systems involved, then decide where a language model, search over your own documents, an agent or ordinary software belongs.
Choose the model and where it runs. We work with OpenAI, Claude and Gemini directly or through Microsoft Foundry, Amazon Bedrock and Google Vertex AI, and with privately hosted open models when your data has to stay in-house. We compare the options on your workload for quality, speed and running cost.
Settle who owns what. Prices, stock, availability and customer records stay with the systems responsible for them. We define what the AI may read, what it may draft and what needs a person's approval.
Build a working slice. A small end-to-end version uses a real data source, a real model and a real screen, so decisions rest on evidence rather than a demonstration.

Some decisions need more than a prompt. When you want to forecast demand, classify incoming documents, rank enquiries for your sales team or spot an operational anomaly, a model trained on your own data can do it consistently. Sphere 10 builds that model, proves it on data it has not seen and delivers it in a form your software can use.

What we do

Start from the decision. We agree what the model must predict, what information is available at that moment and what a wrong answer costs you.
Prepare the data properly. We check how past examples were labelled and whether history reflects how the model will be used, so results hold up outside the test.
Pick the right technique. Forecasting, classification, ranking, image recognition and language-model fine-tuning each have suitable tools. We compare candidates against a simple baseline and train a custom model only where it earns its place.
Fit it into your application. Predictions arrive through a documented interface, in .NET or as a service, with uncertain cases sent to a person.

Your staff and customers should be able to ask a question and get an answer drawn from your own manuals, procedures, product information and records, with a link to the source. Sphere 10 builds that knowledge capability into your application, so answers can be checked, access is respected and updates flow through.

What we do

Connect your sources. SharePoint and OneDrive, Notion, shared documents, help content and business records are brought in through supported interfaces, keeping their structure, versions and access rules.
Make answers traceable. Every answer cites the passage it came from. Where the information is missing or contradictory, the assistant says so instead of guessing.
Respect permissions. People see answers only from material they are entitled to read, and deleted or withdrawn documents stop appearing.

Turn documents into records. Invoices, applications and forms can be read into structured fields that your staff review against the original before approving.

What you receive

A working assistant or search feature inside your application, with source links.
Connectors that keep the knowledge collection current, including deletions.
A set of test questions that measures answer quality and catches regressions.
Procedures your team can follow to add sources, rebuild and resolve failed imports.

An agent is useful when it finishes work, not when it only talks about it. Sphere 10 builds agents that take an enquiry, booking, document or order from arrival to a completed record in your systems, with your staff approving the steps that matter and able to see exactly what was done.

What we automate

Lead intake and follow-up. Enquiries from your website, email or SMS become complete CRM records with the next task assigned, and follow-up messages stop when the customer replies or books.
Quotes, agreements and signatures. Approved details populate your templates, and signed documents are filed against the right customer or job.
Orders and dispatch. Order events become preparation, pickup or dispatch queues, with duplicates and incomplete orders flagged for staff.

Bookings and enrolments. The agent checks real availability, gathers the details you need, confirms within your rules and sends reminders that follow changes and cancellations.

How we keep it under control

Each action the agent can take is defined, permission-checked and logged.
Reading, proposing and committing a change are separate steps where your business needs review.
Work survives interruptions. A retried step does not create a second booking, message or payment.
Staff can inspect, pause and take over any run.

AI is most useful inside the software people already use. Sphere 10 adds assistants, document understanding, guided analysis and drafting to your website, web application, desktop software or back-office system, sharing its sign-in, data and controls instead of sitting in a separate chat window.

What we build

Customer assistants. Answer product, service and policy questions from approved information, look up a customer's own order or booking, and hand over to staff with the conversation attached.
Staff assistants. Help your team find the right record, prepare a reply or draft a handover from within their normal working screen.

Storefront assistants. Recommend suitable products from your live catalogue and availability, and answer order questions from the real order.

Content production. Turn approved source material into briefs, articles, newsletters, social posts and campaign imagery, with editorial review before anything is published.

What you receive

A complete, working feature in your product, not a prototype.
Connections to the systems that own your products, customers and orders.
Review and approval steps where your business requires them.
Usage, cost and failure reporting, plus the test examples used for acceptance.

A model that works in a demonstration still has to be run, monitored, updated and paid for. Sphere 10 takes AI from a working build to a dependable service, on a managed cloud platform, dedicated GPU servers or a private environment, with clear ownership of data, cost and day-to-day operation.

What we do

Size it for real use. We measure response time, throughput and cost with representative requests, then choose hosting and hardware from those measurements.
Release safely. Model and software versions are recorded, tested against agreed examples before release, rolled out in stages and reversible.
Monitor what matters. Alongside system health, you see whether the business work completed, including unanswered enquiries, pending approvals, failed imports and running cost.
Keep changes controlled. Updates to your policies, prices, documents and templates are traceable and reach the service without a redeployment.