Frequently asked questions
Short, direct answers. If yours is not here, ask us.
About Ollie Solutions
What does Ollie Solutions do?
Ollie Solutions builds AI voice agents, chatbots, automations, machine learning models, websites and custom software for businesses, trains their teams to use AI, and helps them get found in AI search. Projects are fixed-scope and start with a free conversation, by call or by message.
Who do you work with?
Online businesses and SaaS companies anywhere, and local businesses in Vancouver and across British Columbia.
Who has Ollie Solutions worked with?
We are the creators of OllieML (ollieml.com), a consumer AI app with its own machine learning model. We built the MCP server, agents and video-classification model for BrandSearch, and we lead the software and machine learning for the Artisan Bionics prosthetic hand.
How much does a project cost?
Every project is quoted as a fixed price after the first conversation, once the scope is written down. You know the full cost before any work starts.
How do we get started?
Book a free 20-minute call or send us a message, whichever you prefer. You describe how the work gets done today. We point to the two or three places where AI or software would save the most time or money, say roughly what each would take, and tell you honestly if something is not worth automating.
How fast can you start?
Most projects start within two weeks of the first conversation, and the first working version is in your hands within the first two weeks of the build.
AI voice agents and receptionists
What is an AI receptionist?
An AI receptionist is a voice agent that answers phone calls for a business. It greets the caller, answers questions from information you provide, books or changes appointments, and passes the call to a person when it cannot help.
Will callers know they are talking to an AI?
Yes. The agent says so at the start of the call. Modern voices sound natural, and being upfront keeps trust with your customers.
What happens when the agent cannot answer?
It takes a message or transfers the call, depending on the rules you set. Nothing is guessed: if the answer is not in what you gave it, it says so and hands off.
Affordable AI chatbots for customer support and sales
How is an AI chatbot different from the old scripted chatbots?
Scripted chatbots follow a fixed menu. An AI chatbot understands questions in plain language and answers from your own documents and data, so it can handle questions nobody scripted in advance.
How do you stop a chatbot from making things up?
It answers only from the information you give it, shows where the answer came from, and says it does not know when the answer is not there. We test it against real customer questions before launch.
Where can the chatbot run?
On your website, in text messages, on WhatsApp, in Messenger or inside your own app. One chatbot can serve all of them.
AI automation and integration
How do I integrate AI into my business?
Start with a list of tasks that are repeated, rule-based and currently done by hand. Rank them by hours spent. Automate the top one or two with AI connected to the software you already use, measure the result, and only then move to the next.
Will this replace our current software?
No. Integration means AI works through the systems you have. Replacing software is slower, riskier and usually unnecessary.
Is our data safe?
We use providers that do not train on your data, keep access to the minimum each integration needs, and can run models on your own infrastructure when the data must not leave it.
AI agent development
What is the difference between an AI agent and a chatbot?
A chatbot answers messages. An agent does work: it is given a goal, chooses its own steps, calls tools such as your database or email, checks the result and continues until the job is finished or it needs a person.
How do you stop an agent from doing something harmful?
By design, not by hoping. Agents get read-only access by default, a short list of allowed actions, hard limits on spending and volume, automatic checks before anything is published, and a log of every step.
Can an agent run without expensive model costs?
Often, yes. Most of an agent is ordinary code. We use small or free models for routine decisions and reserve larger models for the few steps that need them.
Websites, apps and custom software
What do you build with?
Mostly TypeScript, React and Next.js on the front end, Python and Node on the back end, Postgres for data and Cloudflare or your cloud of choice for hosting. We pick boring, well-supported tools.
Do we own the code?
Yes. You get the repository and every credential. Nothing is locked to us.
Can you take over an existing project?
Yes. We start with a short review of the code and tell you plainly what should be kept, fixed or rebuilt.
Machine learning model training
When should we train a custom model instead of using ChatGPT or Claude?
When the task is narrow, high-volume and needs to be fast or cheap. A small model trained on your examples can classify thousands of items a second for a fraction of the cost. For open-ended reasoning or writing, a large model is still the better tool.
How much data do we need?
Less than most people expect. A few hundred labelled examples is often enough to fine-tune a small model for classification. If you have none, we start by labelling a sample with you.
Have you built models that run in production?
Yes. OllieML's own face-matching model compares an uploaded photo against thousands of celebrities and serves the public site at ollieml.com. We also built the video-classification model for BrandSearch and the grip-recognition neural network for the Artisan Bionics prosthetic hand.
MCP server development
What is an MCP server?
An MCP (Model Context Protocol) server is a small service that describes your product's actions to an AI model in a standard format. Once it exists, any assistant that supports MCP can search your data and perform actions on a user's behalf, with their permission.
Do we need a public API first?
No. If you have one, the MCP server sits on top of it. If you don't, we build the server against your database or internal services and add the access controls an external API would need.
How long does an MCP server take to build?
A focused server with a handful of well-designed tools usually takes two to four weeks, including testing with real models. Scope is fixed after the first conversation.
AI search optimization (GEO)
What is generative engine optimization (GEO)?
GEO is optimizing a website so that AI assistants quote and recommend it. It focuses on clear factual answers, structured data, consistent brand information and being mentioned on sources the models already trust.
How is GEO different from SEO?
SEO aims for a high position in a list of links. GEO aims to be the source an AI uses when it writes one answer. Much of the groundwork is shared, but GEO puts more weight on direct answers, citable facts and brand mentions than on keywords and backlinks.
Does it work?
It does for us. ChatGPT is the largest single source of visits to our own product, ollieml.com, ahead of Google.
AI training for your team
Do we need technical staff for AI training?
No. The training is for people who do the everyday work: admin, sales, marketing, support and management. No coding is involved.
Is it safe to put company information into AI tools?
It depends on the tool and the plan you are on. Part of the training is a plain-language policy: which tools are approved, and what information must stay out of them.
How long is a workshop?
Usually a half day, in person or online, followed by a shorter session a month later to fix what did not stick.