AI case notes for social workers: on-device documentation for home visits and safeguarding

Social workers in England spend — depending on the survey you cite — somewhere between 60 and 80 percent of their working time on administration. Case recording, assessment documentation, referral paperwork, court reports, safeguarding planning records. The time not spent on administration is the time spent with the people the service exists to support.

This ratio is the central crisis of UK social work. It is not a new observation. The Munro Review of Child Protection (2011) identified excessive bureaucracy as a structural threat to good practice. Fifteen years later, the ratio has barely shifted.

AI is beginning to change it. But the tools in widest use carry their own risks — and for social workers working in the field, those risks are not hypothetical.

The Magic Notes pattern — and its limits

Magic Notes by Beam is the most widely deployed AI documentation tool in UK social care, currently used by over 85 local councils. It records home visits and team meetings, transcribes them, and generates structured case notes aligned to the format each council uses. Social workers report significant time savings — some citing an hour or more per day — and improved quality of direct engagement, because attention previously split between listening and writing can be fully on the service user.

The evidence for Magic Notes is real and the benefit to practitioners is well-documented. Two things are also true:

Council procurement gates access. Magic Notes is deployed at the council level, not the individual level. A social worker at a council that hasn't procured it cannot use it. Many councils are still evaluating. Procurement cycles in local government are slow. A newly qualified social worker at a council that will deploy Magic Notes in 18 months has 18 months of unchanged admin burden in the meantime.

It's cloud-based. Audio from home visits and team meetings travels to Beam's servers for processing. Service user data — which routinely includes health conditions, mental health history, domestic abuse disclosures, substance use, sexual orientation, and child protection concerns — is Article 9 special category data under UK GDPR. The DPA 2018 and UK GDPR require the highest level of protection for this data category. Transmitting it to a cloud processor creates a data trail that a council-level DPIA must address.

For social workers whose council hasn't deployed Magic Notes, or who work in contexts where that cloud exposure is a compliance concern, a different architecture is available.

On-device processing: what it changes

Kuulo records home visits, assessment meetings, and team discussions on an iPhone, processes the audio entirely on-device, and generates structured case notes without transmitting audio to any server.

Service user disclosures, child protection concerns, safeguarding discussions — all processed locally. The data protection question shifts from "what cloud processor receives this audio and under what DPIA?" to "what happens on this practitioner's device?" That is a materially simpler governance question.

For individual social workers who want to use AI for case recording without waiting for council procurement, on-device processing removes the central data governance barrier.

Home visits and no-signal environments

Social workers conduct the most sensitive and most important parts of their work in people's homes, in community settings, in hospitals, and in temporary accommodation. These environments are frequently without reliable mobile data. A tool that requires internet connectivity to transcribe or summarise is a tool that fails at exactly the moments documentation matters most.

Kuulo runs entirely on the Neural Engine of an iPhone. It does not require signal to transcribe, summarise, or generate structured case notes. A social worker conducting a Section 47 enquiry in a flat on the fourteenth floor of a tower block — where mobile data may be absent — has the same documentation capability as one in a council office with full WiFi.

The offline capability is not a convenience feature. For social workers, it is a functional requirement.

Children's social work

Children's services documentation has specific requirements:

Home visit records must capture the child's presentation, the home environment, interactions observed, the parent or carer's engagement, and the social worker's professional judgement. The quality of this record is the quality of the safeguarding case.

Child protection conferences involve multiple professionals — police, health visitors, teachers, parents, the social worker — and generate complex information streams that a single practitioner documenting retrospectively will not capture completely.

Child's voice documentation — recording what a child said, in their own words, about their experience — is a specific legal requirement in many contexts. "In the child's own words" is the standard. Paraphrasing from memory hours after the visit is a weaker standard than a transcript of what was said.

Kuulo's diarization separates voices in a multi-participant meeting. A child protection conference recording can produce an attributed transcript in which the health visitor's concerns, the school's observations, and the parent's response are separately identified — a significantly more useful record than undifferentiated notes.

Adult social care

Adult social care assessments have their own documentation demands:

Safeguarding enquiries (Section 42) require detailed records of what the adult disclosed, what was observed, and what decisions were made and on what basis. Retrospective recall is a weaker evidentiary standard than contemporaneous recording.

Care Act assessments require structured capture of the adult's needs, wishes, and the social worker's judgement about eligible needs. The assessment is a legal document that drives care provision decisions.

Mental capacity assessments require two-stage documentation — the decision to be made, and the assessment of capacity for that specific decision — which benefits from structured template output rather than freeform retrospective notes.

Kuulo's Care Assessment template structures output across: presenting concerns, service user's own account, observed environment, risk factors identified, professional judgement, and recommended actions. This maps onto the structure of most council recording formats.

The court bundle

Social work records become legal evidence. Child protection proceedings, Court of Protection cases, and tribunal hearings all draw on the case records that social workers create. The quality of those records is material to the legal outcome.

A court bundle drawn from contemporaneous, structured, accurately attributed case notes is a different quality of evidence from one drawn from retrospective summaries written hours or days after the events they describe. Judges notice the difference. Lawyers notice the difference.

The argument for AI-assisted case recording is not primarily an efficiency argument — though the time savings are real and significant. It is also a quality argument: the records produced from contemporaneous AI-structured notes are more accurate, more complete, and more defensible under cross-examination than records written from memory at the end of a long shift.

For newly qualified social workers

NQSWs (newly qualified social workers) in their Assessed and Supported Year in Employment face the double burden of developing practice competence while also meeting the administrative demands of a full caseload. Admin time for NQSWs is often cited as even higher than the profession average, because they are slower at writing and less confident in their recording format.

Kuulo's templates provide a structure that NQSWs can record into from their first home visit. The output is formatted correctly, it's complete, and it can be reviewed with a supervisor as a contemporaneous record of the visit. The ASYE portfolio — which requires documented evidence of practice — benefits from recording that is accurate and attributed rather than reconstructed.

Data governance in practice

The practical data governance position for a social worker using Kuulo:

Service user audio is processed on the social worker's council-issued or personal device. No audio is transmitted to a server. The GDPR question is about device security and the practitioner's own obligations — consistent with the council's existing policies on device use and data security.

This is the same architecture a social worker uses when they write a handwritten note in a service user's home: the data is on a physical medium in the practitioner's possession, subject to the council's device and records policies. Kuulo is that handwritten note, structured, complete, and searchable.

Councils that have deployed Magic Notes have done the DPIA work for that tool. Councils that haven't can't extend that coverage to individual practitioners using it independently. On-device processing is the architecture that allows individual adoption without a council-level DPIA.