Look at the person. Lastly drafts the form.
Documentation software for social work and care. Lastly listens to the conversation, drafts your form from the person's own words and waits for your approval. Nothing leaves your organisation.
Book a demo
Made for people whose work is talking with people
Listens, notices, drafts. You decide.
Listens, sentence by sentence
Speech becomes text as you talk, stamped with the time and who spoke, so the person's own words stay theirs. Audio is never stored.
Notices what matters
A thought, a feeling, a strength, a wish. Each appears as a card with the exact words marked and the field it belongs in. One tap adds it as a suggestion.
Drafts, then waits for you
Lastly drafts the whole form from notes, recordings and earlier forms, with a source quote for every suggestion. You accept or reject each one.
From hello to an approved form
Pair your phone
Have the conversation
Tap what matters
Review and approve
Point the phone's camera at the code on your desktop. No accounts, codes or settings to type.
Tell the person, start recording and put the phone on the table. Each line shows the time and who spoke.
Cards appear as things are said. One tap sends the exact words to the right field as a suggestion. Anything else goes to your private scratch pad.
Check each suggestion against its quote and accept or reject it. Hints and a meter show if your text fits the printed form. Then export the PDF.
From hello to an approved form
Four steps, and your eyes stay on the person for most of them.
Point the phone's camera at the code on your desktop. No accounts, codes or settings to type.
Tell the person, start recording and put the phone on the table. Each line shows the time and who spoke.
Cards appear as things are said. One tap sends the exact words to the right field as a suggestion. Anything else goes to your private scratch pad.
Check each suggestion against its quote and accept or reject it. Hints and a meter show if your text fits the printed form. Then export the PDF.
AI that stays where you are
Lastly uses three kinds of model, and each one runs on hardware your organisation controls. Nothing about a person is sent to an outside AI service.
Your phone
Your computer
Your model server
Outside AI services
- Listening
Speech becomes text, sentence by sentence
A speech model tuned for Danish conversation writes each sentence in about a third of a second on your own GPU. Audio is held in memory only and never stored.
Runs on your server
- Deciding
Chooses among options, never writes words
It scores fixed options, such as a sentence's field or the usual next step. It answers in under a second per sentence, cannot invent words, and you always decide.
Runs on your server
- Drafting
Drafts the form, with a quote for each
A writing model drafts the form from notes, recordings and earlier forms. Every suggestion shows its source quote, and you accept or reject each one.
Runs on your server
The models do not learn from your cases.
The models need no internet once installed.
Nothing about a person is sent to an outside AI service.
You choose which models run.
Your data stays inside your organisation
Your desktop is the hub. No cloud service is part of the setup, so you always know where a person's data is.
A local database
Cases, forms and transcripts are stored on the desktop, not on someone else's server.
Straight to your phone
Phone and desktop talk directly over your local network, encrypted and locked to your own desktop. No cloud relay in between.
AI on your own server
Listening, drafting and suggestions run on a server or machine you control, with speech tuned for Danish conversation.
Every change on record
Who changed what, and when, in an append-only audit trail that shows if anything was altered.
Lastly adapts to you, so you only adopt once
Lastly is set up around your forms, your methods and your words, so your team does not have to learn its own.
- 01
Your forms
Existing PDF forms become fillable, and the finished form is exported as a PDF in the layout your team already knows.
- 02
Your methods and guidance
Suggestions and writing hints follow your organisation's way of working, not a method built into the software.
- 03
Your words and roles
Your own vocabulary throughout, with separate roles for professionals and managers.
- 04
What comes next
Lastly suggests the step that usually follows, with a plain reason. If something points to serious risk to someone's safety, it gently suggests talking to a colleague or manager. You decide.
What Lastly never does
Sends a person's data to the cloud.
Keeps the audio of a conversation.
Puts anything in a form without your approval.
Makes a judgement about a person.
Blurs the person's words with yours.
Hides who changed what, or when.
Getting started
Three steps from a first look to using it in your own casework.
- 1
See it with a form like yours
Book a demo and bring a form you use. Bring your data protection lead too, if you like.
- 2
Set it up around your work
Lastly is set up with your PDF forms, methods, vocabulary, roles and steps, with the speech and writing models on a server your organisation controls.
- 3
Start small
You can start small: a few colleagues pair their phones and try it in their own casework.
Questions you might have
Where is our data stored, and does any of it go to the cloud?
It is stored in the database on the Lastly desktop app, on a machine your organisation controls. Transcription, drafting and suggestions also run on your own server. Nothing about a person goes to the cloud.
Does Lastly need an internet connection?
Lastly uses no cloud service. Phone and desktop need the same local network, and the desktop needs to reach your own model server, which can sit on that same network.
Does the AI learn from our cases?
No. The models are used as they are and do not train on your cases. Nothing about a person is sent to an outside AI service.
Can the AI make things up?
The model that suggests fields and next steps only scores fixed options, so it cannot invent words. You always decide. The drafting model does write text, so every suggestion shows its source quote and waits for your approval.
Which devices and hardware do we need?
A computer for the desktop app, a phone or tablet for recording, and a server or machine your organisation controls for the models. Ask us about the setup in a demo.
Is the person told about the recording?
You choose when to record, and you tell the person first, following your organisation's guidance. Only text is kept, marked with who spoke. The audio is never stored.
What if Lastly mishears something?
Every suggestion shows the exact words it came from, so you can check it before you accept it. Nothing goes in a form without you.
Can Lastly fill in a form on its own?
No. Every suggestion waits for a professional to accept or reject it.
Which languages does it support?
Lastly is built in Denmark and works in Danish first, with speech recognition tuned for Danish conversation. Ask us about other languages in a demo.
Can we keep our forms?
Yes. Your PDF forms become fillable, and the finished form is exported as a PDF, so you can file it where you do today.
What does it cost to try?
Book a demo and we will walk you through what a start could look like for your organisation.