Most businesses do not need another empty AI chatbot account.
They need customers to get reliable answers without waiting for office hours. They need fewer repeat questions reaching the team. They need the knowledge already sitting in websites, help centres, FAQs and product docs to become easier to use. And they need all of that without turning a support improvement into another internal implementation project.
That is the product lesson behind Solviro, the managed AI support assistant now run by BPS Designs.
Solviro is deliberately not positioned as “paste in a script and hope for the best”. The product is built around approved business content, setup by our team, answer testing before launch, company-level separation and a simple website chat experience. That sounds less dramatic than a demo where an AI assistant confidently answers anything. It is also far closer to what small teams actually need.
The useful question is not “can AI answer customer questions?” It can, sometimes very well. The better question is: what does the product need around the model so the answers are controlled, current and useful?
The weak point is usually the knowledge, not the model
Customer support automation tends to fail in boring ways.
The assistant cannot find the current delivery policy. It mixes up old and new pricing. It answers from a page that nobody meant to expose. It gives a confident reply when the right behaviour would be to hand over to a person. It works during the sales demo, then becomes another system nobody owns.
Those problems are not solved by a bigger chat window. They are knowledge operations problems.
For a support assistant to be genuinely helpful, the business needs to know:
- which public sources the assistant can use
- which pages have been reviewed and approved
- which sections should be ignored
- what fallback message should appear when the answer is not available
- who can change source content, settings and access
- how conversations will be reviewed after launch
- when a human should step in
Solviro makes those concerns part of the product rather than leaving them as an exercise for the customer. It connects public sources such as websites, help centres, product documentation and FAQs, then keeps approval in the workflow before content is used. If Solviro cannot find the information in approved content, it uses the chosen fallback instead of pretending.
That distinction matters. Good AI support is not just generation. It is retrieval, boundaries, approval and review.
Managed setup is a product decision
Plenty of SaaS tools treat setup as the customer’s problem. Sign up, connect sources, configure the widget, test prompts, invite the team, troubleshoot the first install, then work out whether the answers are safe enough to put in front of customers.
That can be fine for technical teams with spare capacity. It is a poor fit for many small businesses.
Solviro takes a managed route. The setup process starts by learning what customers ask, where the answers live and when a person should be involved. Then the useful public sources are connected, irrelevant sections are left out, the assistant is configured, and real customer questions are tested before launch.
This is not just a service wrapper around software. It changes the product shape.
When setup is managed, the product can assume that important choices will be made deliberately:
- the first knowledge sources are chosen with the business, not guessed by a wizard
- the greeting and fallback can match the company’s support style
- approved website domains can be checked before the widget goes live
- sample answers can be reviewed before customers see them
- the installation step can be handled without making the customer interpret technical docs alone
For BPS Designs, that is a product-led decision. The value is not only the AI answer. The value is the operational path from “we get too many repeat questions” to “customers can get useful answers on the website”.
Approved content keeps trust in the loop
The phrase “AI support” can make people imagine a free-roaming assistant that invents its way through every question. That is not the right pattern for most business websites.
Customers usually ask grounded questions:
- How quickly can you deliver?
- Which areas do you cover?
- What is included in the plan?
- How do I reset my account?
- What happens after I book?
- Can I speak to someone if this does not fit?
The business often already has the answers. The problem is that the answers are scattered across pages, FAQs, product copy and help articles. Customers do not want to search through all of that, especially on mobile. They want to ask the question in their own words and get the relevant answer.
That is where approved content is the right constraint.
Solviro is designed to answer from reviewed business content, show the team which pages were used, and keep new or updated pages out of use until an authorised administrator approves them. That creates a healthier support loop: customer questions reveal missing or unclear information, the team improves the underlying content, and the assistant becomes better because the source material is better.
The assistant is not a replacement for content quality. It is a reason to take content quality seriously.
Conversation history is product feedback
One of the more useful parts of AI support is not the individual answer. It is the pattern of questions that appears once customers have a low-friction way to ask.
Conversation history can show:
- the same pre-sales objection appearing repeatedly
- a support article that exists but is hard to find
- a pricing detail that is technically present but unclear
- confusion caused by wording in the product or website
- gaps between what the company thinks customers know and what they actually ask
That makes support automation part of product learning.
If a customer asks a question and Solviro uses the right source, the business learns that the content is doing useful work. If customers keep asking something that has no approved answer, that is not just a support failure. It is a content, onboarding or positioning gap. If the fallback appears too often, the knowledge base needs attention.
This is why the admin experience matters. Teams need to see conversations, source usage and missing information without digging through raw logs. They need enough visibility to improve the product surface, not just admire a dashboard.
Security and separation are part of the offer
Customer support conversations can contain sensitive commercial details, contact information and operational context. Even when the assistant is only using public website content, the surrounding system still needs sensible boundaries.
Solviro’s public security positioning focuses on company-level separation, controlled account creation, approved domains, database safeguards, role-based team access and established providers for identity, hosting, billing, database and AI services. That is exactly the right type of information for a product like this to surface early.
Small businesses may not ask for a full security questionnaire before trying an AI support assistant. But they still need to know that their content, users, assistants and conversations are not being blended with someone else’s. They need to know who can access the account. They need confidence that a website widget will only run where it is meant to run.
Security is not an afterthought bolted onto support automation. It is part of why the product can be trusted enough to put in front of customers.
The real product is the operating model
Solviro is a useful example of a broader principle for AI products: the model is rarely the whole product.
The product is the operating model around the model.
For AI support, that means:
- a clear intake process for customer questions and source material
- reviewed knowledge sources rather than an uncontrolled content scrape
- answer testing using the questions customers really ask
- fallbacks that admit when information is not available
- team roles so the right people can manage content, settings and conversations
- conversation review so the assistant improves the business, not only the response time
- sensible technical boundaries around company data and website access
That operating model is less glamorous than a prompt. It is also much more defensible.
It turns AI from a novelty into a support capability. It gives the business a way to reduce repetitive work while keeping humans responsible for the knowledge, tone and decisions that matter.
What this says about BPS Designs
BPS Designs has been shifting towards building and operating its own software products. Solviro fits that direction neatly because it is product-led, operationally grounded and focused on a real small-business problem.
The point is not to chase AI as a label. The point is to package a useful capability in a way that customers can adopt without needing to become AI implementation specialists.
That is where small product companies can create genuine value. Not by adding a chat interface to everything, but by doing the careful work around it:
- simplify the workflow
- make setup realistic
- create sensible defaults
- expose the controls that matter
- keep evidence visible
- design the support loop after launch
Those are craft decisions. They are also business decisions.
A better way to judge AI support tools
If you are evaluating an AI support tool, do not start with the flashiest answer it can produce. Start with the system around the answer.
Ask:
- What sources can it use?
- How are sources reviewed and approved?
- What happens when the answer is not available?
- Can the team see which content was used?
- Can common questions feed back into better support content?
- How are users, roles and company data separated?
- Who owns the setup and launch?
Those questions will tell you more than a polished demo ever will.
AI support is useful when it helps customers get accurate answers sooner and helps the business learn from what customers ask. That requires more than a chatbot. It requires managed knowledge, clear ownership and a product that respects the messy reality of support work.
That is the space Solviro is built for.