Ask Fondaro

Ask Fondaro is the chat workspace available throughout your dashboard. Open Ask at the top of the sidebar, choose an existing conversation or start a new one, then type or dictate what you need.

Each conversation is private to your user account. You can reopen, rename, or delete it from the conversation list.

When you open Ask Fondaro for the first time in a browser tab, it continues your newest conversation if that conversation was active within the last 30 minutes. After a longer gap, it starts a fresh conversation using your preferred model. Your older conversations remain available in History, and opening a specific conversation always takes you to that conversation regardless of its age without changing your preference.

When Fondaro will ground the next turn with a specific record, the empty composer names it in plain language: Ask about this lead, Ask about this call, or Ask about this listing. Other views keep the general prompt because they select the relevant capabilities rather than sending every field visible on the page. This context does not grant extra access: your normal role and record permissions still apply.

Who can use it, and what it can see

Ask Fondaro is available to everyone in your organization, not just admins.

It works with your own signed-in identity, so it sees exactly what you see everywhere else in Fondaro. If you are a member, that means the leads, deals and tasks assigned to you. If you are an organization admin, it means the whole organization.

A few things are admin-only and simply will not appear if you are a member: changing who owns a lead, changing the status of many leads at once, creating or editing Loops, and reading or building the organization website. Changes of any kind also need an active subscription, so without one Fondaro AI can read but not propose.

Your identity and branding

Every Fondaro AI turn receives your organization's current branding and the signed-in team member's identity before the model runs. Customer-facing copy uses your agency name, description, languages, contact details, logos, and the current rep where those values are configured. Fondaro is the software provider and is never used as a fallback agency, sender, team name, or email sign-off.

An email proposal resolves its envelope on the server rather than asking the model to invent it. The review card shows the exact From and To email addresses, the rep's name and avatar, and the organization behind the sender. Fondaro AI will not create an email proposal unless the lead has an email address and your organization has a verified sending domain, and it refuses customer email copy that presents Fondaro as the sender, agency, team, or sign-off.

What Fondaro AI can do

The assistant can now work across most of Fondaro:

Leads. List and search them, read a full brief, read the whole activity timeline, count the pipeline, and read your tag catalogue and team roster.
Search by what people said. Find the people whose notes, calls and emails match a plain-language description, and see the evidence behind each match. There is a section on this below.
Lead changes. Create a lead, correct contact details, move a lead's status, and set tags. Admins can also change who owns a lead and move many leads at once.
Tasks and notes. List them for a lead, add one, edit one, delete one.
Deals. The whole pipeline: list and open deals, create one, edit it, move it between stages, close it as won or lost, and reopen it.
Calls and emails. Read a lead's call history, the stored transcript and AI analysis of a specific call, and the email history. It reads only what has already been generated, so asking for a transcript never starts a new transcription.
Properties. List your connected sources, resolve a place name, search listings, inspect a visual contact sheet, and refine the results into a buyer-specific shortlist.
Brochures. The full lifecycle: create one, edit its details, add, remove, reorder and refresh listings, correct a listing's location or facts, revoke and renew the share link, and delete a brochure.
Documents. Read the organization library, open a document, and see what is attached to a lead. It can also write a new markdown document, edit one, publish or revoke its public link, and attach it to a lead or detach it. Every one of those changes is a card you approve first, publishing included.
Reports. Read the metric catalogue, build a saved CRM report, and freeze a snapshot so the numbers it quotes are the numbers you see.
Dialling. Queue an ordered list of leads for a power-dial run. Fondaro AI never places a call; your own browser does that.
Loops. Admins can read the node catalogue and create or update Loops.
Website. Admins can read the current site and preview state, then prepare a brand takeover, ordinary change, language addition, or failed-preview repair. The build starts only after Build preview is approved.

Reads run immediately and come back as compact cards. Every card links to the page that owns the record, so you can go straight there when you need the full picture.

Property searches also give you two direct next steps. You can refine the results into a shortlist, or ask Fondaro AI to prepare a brochure and email. A broad first page is treated as discovery rather than as a recommendation: if the buyer, budget, bedrooms, or another important criterion is missing, the assistant asks for that before claiming a property is a strong match. Generic brochure requests remain editable proposals. During completed-call review only, the grounded brochure is created automatically; the note, status and email stay pending, the email already contains the real link, and an explicit refinement updates that same brochure.

The Match leads action on an owned or external listing opens a focused dialog and provides another direct handoff. Open a matched person's menu and choose an email or brochure action. Fondaro opens Ask on the relevant property context with the person named in the draft request. External source identity travels in the request so the assistant can resolve the listing through the connected source. The resulting email and brochure cards still require your approval.

Try requests such as:

"Show my pipeline."
"Which of my leads have not been contacted in three weeks?"
"Summarise the last call with the Andersson lead and what I promised them."
"Move the Marbella villa deal to negotiation."
"Search my connected sources for three-bedroom villas in Marbella under 900k."
"Build a brochure from those three and set the recipient to Sofia."
"Prepare a follow-up task for that lead next Tuesday."
"Build a Loop that drafts a welcome email when a new lead arrives."
"Change the website homepage to lead with our beachfront listings."

Search your CRM in plain language

Most CRM search only finds what someone typed into a field. Ask Fondaro also searches what people actually said to you, so you can go looking for an intention rather than a keyword.

Describe what you are after the way you would say it out loud:

"Who mentioned a sea view penthouse?"
"Who said they are visiting this summer?"
"Who might like this listing?"

The answer is a ranked list of people, not a wall of text. Each person arrives with the evidence for the match: a short excerpt of what was said, the kind of record it came from, and when it happened. Every row links into the CRM, so you can open the lead and read the whole note, call or email in context.

Order is the only relevance signal. There is no confidence score to read too much into: the person at the top is the closest match Fondaro found, and the excerpts are there so you can judge that for yourself. If two people match equally well, the one you heard it from more recently, and in more than one place, comes first.

Timing is handled precisely. When you say "this summer" or "in the last three months", Fondaro AI turns that into a real date range and the database applies it, so a question about a window gets an answer from that window.

What it searches

The search reads the records you can already open in the CRM:

Your team's notes on a lead.
Stored call transcripts, plus the summary written from each call. Nothing is transcribed to answer a search; it reads what is already there.
Emails a person on your team wrote, and the replies they received. Automated template sends are left out, because they say nothing about the buyer.

It covers the last 24 months of activity. Older history stays reachable through the ordinary CRM filters, lists and name search.

Who it can reach

Your reach is your normal reach, unchanged. If you are a member, the search looks across the leads assigned to you. If you are an organization admin, it looks across the whole organization. Asking a question never widens what you can see, and a person you cannot open never appears in an answer.

What Fondaro AI remembers

Fondaro AI keeps short working notes so you do not have to tell it the same thing twice. These are called Memory, and there are three of them:

Your memory is about how you like to work. "Keep my emails to three sentences." "I work the Estepona pipeline."
Organization memory is what your whole team shares. "We hand Nordic buyers to Elin." "No viewings on Sundays."
A lead's memory is the picture of one person. "The wife decides." "Real budget is closer to 800,000." "Call back on Thursdays."

Memory is not a transcript and not a second set of notes. It is the short summary a good colleague would carry in their head, and Fondaro AI rewrites the whole thing whenever it learns something new rather than adding to the end of it. That is what keeps it short enough to be useful.

Your memory and your organization's memory are read at the start of every conversation. A lead's memory is read whenever the screen you are on is about that lead, including a call review. You do not need to ask for any of it.

See it, change it, clear it

Your memory and Organization memory live on Settings → AI & models. Each one shows the current note, when it last changed, and a History of every version with the person or the assistant that wrote it. Edit rewrites the note, Restore brings an older version back, and Clear removes the note and its history outright. Restoring saves the older version again as the newest one, so nothing is lost by going back.

Organization memory is readable by everyone and editable by organization admins, because it is the team's shared playbook. Your own memory is yours alone: nobody else can read or change it, not even an admin. A lead's memory follows the lead and is used automatically whenever you work with that person.

It also powers search

A lead's memory joins the notes, calls and emails that plain-language search reads, labelled Memory in the evidence. That is how "which buyers dislike stairs?" can find someone even when no note ever said so in those words. Select the person's name in a result to open their lead record.

Your memory and your organization's memory are never searched this way. They are about you and your team, not about a buyer, so they would only ever get in the way of an answer about people.

Limits worth knowing

Each note has a length limit, shown as a counter while you edit. Fondaro AI shortens rather than sprawls when it reaches it.
If two people, or you and Fondaro AI, change the same note at once, the second save is refused and shows you the version that is actually saved so you can fold your change into it. Nothing is silently overwritten.
When two leads are merged, the surviving lead keeps its own memory and the other one is added underneath it.
Nothing read out of memory changes what needs your approval. A note saying "always send the brochure" does not send a brochure.

Review before anything changes

Fondaro AI never changes your data as soon as you ask. It puts a card beside its reply, and nothing happens until you approve it.

There are two kinds of card.

Editable cards for a task, a CRM note, an email, a new brochure, or safe website request fields. These show the fields and let you correct them before approving. An email card keeps its server-resolved From and To identity read-only while you edit the subject and body, and previews the CRM signature that will be appended. Use Edit signature on the card to change it in the existing signature editor. Its action says Send email because approval sends it; the card says Sending until the server confirms delivery. A website card uses Build preview and discloses that approval starts one metered builder turn.

Change cards for everything else. These are a summary of what will change: the record it affects, what it is now, and what it becomes. They are deliberately read-only. If something is wrong, reject the card and ask again rather than editing it, so what you approve is always exactly what you reviewed.

To resolve a card:

Check the record it names, and open it if you want to see more.
Read the before and after values.
Choose Approve to apply exactly what the card shows, or Reject to leave Fondaro unchanged. For an email, choose Send email instead.

Approval runs the same business rules as the equivalent action on the page itself. A proposal cannot be approved twice, and the decision and its result stay beside the original message, including after you close and reopen the conversation. Approved, rejected, and failed cards keep their final state instead of offering the action again. A completed CRM proposal card names the team member who approved it, shows their account avatar, and records Fondaro AI plus the exact model that prepared the turn when that model attribution is available.

Two behaviours worth knowing:

Changes that cannot be undone are marked. Deleting a task, note or brochure, revoking a brochure or document link, and changing the status of many leads at once all carry a warning and a distinct confirm button.
If the record changed while the card was sitting there, approval fails. Rather than overwriting whatever someone else did in the meantime, Fondaro AI tells you it changed since it was proposed. Ask again and it will read the current state.

An email proposal is not sent while it is pending. Choosing Send email rechecks your current access, the lead's current email address, and the reviewed organization sender. If they still match, Fondaro sends it through your verified domain, applies your CRM email signature, and records it in the lead's email history and timeline. If anything changed, sending fails and you must prepare a fresh proposal.

Choose a model

Your first conversation starts with Auto, the recommended choice. You can choose a different model at any point. Fondaro starts a new conversation with that model automatically, saves it as your preference for this organization, and keeps the previous conversation unchanged in your history. New conversations started elsewhere in the dashboard and background call reviews use the same preference, including after a cold browser load. Choose Auto to make Auto your preference again.

Auto is deterministic. It uses the first available model in Fondaro's approved list, and the model control keeps that model's name visible beside Auto. It does not balance a turn dynamically between models.

The searchable selector groups the approved models into Flagship, Balanced, Fast, and Free sections. Each row shows the provider, model name, an input capability summary, and Fondaro's effective input and output prices per one million tokens in your billing currency. These are the customer rates after Fondaro's configured pricing has been applied, not the provider's base cost. Cached input and cache creation have their own prices when the live catalog supplies them.

A Free label appears only while every live token price for that server-designated free model is exactly zero. If any of those rates changes, Fondaro removes the model rather than presenting a paid model as free.

Token rates are displayed as standard two-decimal currency values, while the usage ledger and payment meter store whole cents. Each model step with a positive cost rounds up once to the nearest cent. The selector shows this rule beside the rates. If live prices cannot refresh temporarily, it labels the last saved snapshot as stale; it never estimates a replacement rate in the browser.

Models can accept different inputs. DeepSeek V4 Flash and Laguna are text-only, so neither can be selected while a draft or the conversation contains images. When either model is selected, the composer blocks image selection, drag and drop, and pasted images. The server checks the same capability again before any provider call.

DeepSeek V4 Flash and Laguna are the models for which Fondaro does not enforce a no-training provider route. Their May train labels explain that DeepSeek or Poolside may use the conversation and CRM context for model training. Fondaro keeps the no-training safeguard enabled for every other model. If Auto ever resolves to a model with the same policy, Auto shows the same label.

When Auto is selected, each answer names the model that actually produced it. Explicitly selected models are named on answers when a conversation contains more than one model, so you can tell them apart later.

The available list comes from Fondaro's live model catalog, so names can change as models are evaluated or become unavailable. If a model is not covered by your plan, the reply is replaced by a short note naming that model and your conversation is kept. Choose another model and send again.

If an explicitly selected model is temporarily busy or unavailable, the failed-turn note can offer Switch to Auto when Auto currently points to a different model. Choosing it starts a new conversation with Auto and keeps the failed conversation unchanged in History. You can instead wait and retry the original conversation.

Dictate a message

Choose the microphone in the composer and record your request. Fondaro turns the recording into text in the composer. Review or edit that text, then send it like any other message. This is dictation, not a voice call or a spoken assistant session.

Background work

Most turns finish inside the open conversation. Work that genuinely detaches is followed on the surface that owns it: website builds on Website, CRM exports on Migrations, and video renders on the ad you started them from.

Website work is durable after approval. Closing the dock or navigating away does not cancel it. Its compact work card restores a safe phase when you return, while the Website preview remains the source of truth. See Build your website with Fondaro AI.

Ordinary answers, reads, and proposals are not background work. Keep the conversation open while a response is generating. If a turn is interrupted, use Retry or Regenerate. Use Stop when you want to end the current response.

Current limitations

Fondaro AI can only use the capabilities listed above. It does not have unrestricted access to every Fondaro screen or API operation.
It does not send email without your explicit approval, dial a phone, publish a website, or approve its own proposals.
Long call transcripts and email bodies are shortened before the assistant reads them. When the exact wording matters, open the record.
Plain-language search finds and ranks people; it does not count or total them. Ask Reports for "how many" questions.
A person is matched on what is written in their notes, calls and emails. Someone whose preferences were never written down anywhere cannot be found this way.
A change card cannot be edited. Reject it and ask again.
Fondaro AI can queue a dial session, but it cannot pause, resume, advance or stop one. Use the dialer's own controls for that.
Dictation only fills the text composer.
Loop authoring is admin-only. Fondaro AI creates a new AI-authored Loop paused for review and links you to its editor.
Loop runs are followed in the Loop editor.
Website building is admin-only, creates a preview rather than publishing, and allows one active builder turn per organization.