Professional services · website AI assistant
AI chatbot for Professional services
Updated
The short version
Answer scope, process, and fee questions from your own published material — and qualify the enquiry before it reaches a partner's calendar.
Firms — legal, accounting, consulting — sell judgement, and judgement doesn't scale. But the questions that arrive before the judgement does are remarkably repetitive: what do you cover, how do you charge, how long does it take, what do I need to bring to a first meeting.
Handling those by email costs senior time and answers slowly. Handling them badly costs the enquiry.
MySiteGPT answers them from your service pages, guides, and published fee information, with a citation on every claim, and captures a name and email when someone is ready to talk. What it will not do is give advice — anything beyond your published material returns your fallback, which is the only acceptable behavior when the alternative could be read as professional advice.
What professional services firms get asked
Real questions, in visitors' own words. Every one of these is answerable from content you probably already publish.
- “Do you handle this type of matter?”
- “How do you charge — fixed fee, hourly, or retainer?”
- “How long does this process usually take?”
- “What documents should I bring to a first meeting?”
- “Do you work with clients in my sector or region?”
- “Do you offer a free initial consultation?”
How MySiteGPT helps professional services firms
Qualifies enquiries before they cost senior time
Scope and fee questions get answered from your published material, so the enquiries that reach a partner are the ones worth their hour.
Never strays into advice
It answers strictly from your content. Anything that would require applying judgement to someone's facts hits the fallback and becomes a captured enquiry instead of an opinion.
Your published material, finally read
The guides and insight pieces your team writes get surfaced as the answer to a live question, with a link — rather than sitting in a resources archive nobody browses.
Enquiries land while intent is high
Lead capture takes the name, email, and the question asked, and alerts your inbox immediately — the difference between a callback and a competitor.
What to train it on, in order
The answers are only as good as the index behind them. Start at the top — these are ordered by how much question volume they absorb per page trained.
- 1
Service and practice-area pages
Defines what you do and don't cover, which is the first question every enquirer has.
- 2
Published fee structures and engagement terms
Train only what you've actually published. Where the answer is 'it depends', say that in a Q&A pair rather than leaving it to retrieval.
- 3
Guides, insights, and FAQ content
Your existing thought leadership is the substance behind good answers — and each citation sends the reader to a page that demonstrates expertise.
- 4
Onboarding and first-meeting checklists
Practical, high-volume questions with stable answers, often only available as a PDF. Upload it.
Not sure your pages are ready to be retrieved from? Run one through the RAG readiness checker first, or audit the whole site with the sitemap analyzer.
What to measure once it's live
We don't publish other people's numbers as if they were yours. Track these instead — they are all visible in your own dashboard within a fortnight.
- Enquiries captured per month, and how many were in scope
- Senior time spent on pre-qualification calls
- Fallback rate on advice-shaped questions — high is correct here
- Unanswered questions naming services you offer but haven't written up
AI chatbot for Professional services: frequently asked questions
Could it accidentally give legal or financial advice?
It composes answers only from passages of your own published content and cites them, and returns your fallback when retrieval confidence is low. Write the fallback to say plainly that the assistant provides general published information and cannot advise on individual circumstances, and keep unpublished internal material out of the training set.
Can we control the exact wording of sensitive answers?
Yes — that's what Q&A pairs are for. Anything your compliance or risk function needs to phrase precisely should be trained as a Q&A pair rather than left to be summarised from a longer page.
Where do captured enquiries go?
To an instant email alert and the leads table in your dashboard, and to a webhook if you want them in a CRM or matter-management system.
Can a person take over a conversation?
Yes. Live conversations are visible in the dashboard and a human can take over a thread mid-conversation when an enquiry is worth intercepting.
Try it on your own content first.
The chat bubble on our homepage is the product, trained on our own site — ask it anything and watch every answer cite its source. Then train one on yours: 14 days, full features, no card.