Beam’s first customer story came with a question no checkout page could answer. Tony, living in a homeless hostel, was told that strangers would pay for his electrician training. “But I don’t understand. Why would anyone help me?” he asked, according to co-founder Alex Stephany’s account. In the end, 135 people did. Tony qualified, found work and eventually helped keep a hospital running.
It was an appealing proposition: show people a practical obstacle, put a price on removing it and let the public help. Today, the donation button has gone. Beam sells software that writes draft case notes, translates conversations and handles service calls. It also continues delivering housing and employment support funded by government. The route between those businesses runs through an unromantic problem: somebody has to pay for the work around the good deed.
The people behind the donation button
Stephany brought experience running JustPark, the parking business, and an interest in how technology could connect people’s resources. Co-founder Seb Barker brought experience in homelessness and healthcare, including work at Thames Reach, Turning Point and NHS England. Beam’s own history places their meeting in 2016. The following year, they began putting the idea into practice.
The pairing matters. A website can collect money; helping someone move into work requires a different set of skills. Beam’s early referral partners included St Mungo’s and Thames Reach. Members worked with a support professional on a plan and an itemised budget. The list could include travel or childcare alongside a qualification. The awkward little costs were part of the job.
That practical streak survives. Beam’s Brent partnership describes help with CVs, applications and interviews, as well as funding for equipment, training and travel. On the housing side, support can cover a deposit, advance rent and moving costs, with a caseworker helping people find a property and complete the paperwork.
For someone considering the service, this distinction is useful. Beam offers assistance with the route into a job or tenancy. It cannot conjure a vacancy, an affordable flat or a willing landlord. Its work is most relevant where a person can move forward once several manageable obstacles are removed, and where a local service can connect them with the programme.
A business needs more than a good campaign
The financing problem appeared early. In a 2018 account published by Nesta, Beam described funding the organisation as its biggest challenge, despite grants, publicity and promising results. That is the first clear constraint in the public record: financing the team and platform around the campaigns. Raising money for an individual’s course did not settle the company’s operating budget.
The model changed over time. Nesta’s early description allocated 10% of donations to running costs. Beam later operated a 100% giving arrangement with the separate Beam Foundation, which received charitable donations for members’ expenses. Those versions belong to different moments. Neither should be mistaken for the business a council buys from today.
When Stephany announced the end of crowdfunding, he said more than seven years of service evidence had made it possible to move members’ expenses to government funding. Beam reported having helped over 5,000 people into jobs and homes. The foundation would continue separately as a grant-making charity. The company’s argument was that public commissioning could support greater reach than donations.
For founders, the useful lesson is fairly demanding: measure whether the service works, then establish who has both a reason and a budget to purchase it. Changing the payer took years of delivery. A more polished fundraising page would not, by itself, have accomplished that.
The product hiding in the paperwork
By 2023, Beam’s own caseworkers were writing thousands of notes each month, according to its company history. The organisation built an AI tool to help. It became Magic Notes, subsequently renamed Beam Notes as its functions expanded into more elaborate documents and workflows. The first users were colleagues already doing the work.
Swindon Borough Council supplied an early external test. Between April and July 2024, 19 social workers and one leadership support officer used the tool across 184 frontline meetings. The council reported a 63% reduction in the time spent compiling assessments and logging case notes. Average write-up time for Care Act conversations fell from four hours to one hour and 30 minutes.
Same write-up task.
A shorter afternoon.
Care Act Conversation, on average
Those are useful numbers with a defined setting. They describe a limited council pilot, rather than a universal productivity rate. They also measure time, which is only part of the public-service bargain. A faster record still needs to be accurate, and the released hours need to become something worthwhile for staff or residents.
The next decision was concrete: Swindon secured a further six-month contract. That purchase is a more revealing vote of confidence than an enthusiastic adjective. It shows a buyer moving beyond a trial, while leaving the longer-term outcomes open to examination.
Three tools, one crowded working day
Notes starts with a recording or uploaded material and produces a draft in a selected template. A practitioner reviews and refines it before transferring it into the organisation’s system. The attraction is the distance between a conversation and a usable record: staff can spend less time reconstructing what happened after the person has left.
Talk tackles incoming demand. Beam describes voice agents that capture information, prioritise calls and hand conversations to human staff with a summary. Out-of-hours coverage and locally tailored knowledge are part of the offer. The practical buying question is which enquiries a service can safely handle this way, and which require a person immediately.
Interpret provides live translation between two selected languages, with speech and text output and a bilingual transcript. Its development contains a telling reversal. Beam says the first version required tapping to translate; frontline feedback led it to rebuild around automatic speaker detection. Removing the button was an attempt to make the conversation feel less interrupted.
This puts Beam in the market for specialised software for public and human services. Functional alternatives range from manual documentation to general transcription tools and human interpreting. Its strongest distinction is its experience delivering frontline work itself. That gives it a basis for understanding why a fluent summary can still be the wrong document.
A £75,000 line in a council register
The business model becomes visible in procurement. Bath and North East Somerset’s register records a £75,000 awarded value for Magic (Beam) Notes, awarded on April 30, 2025. The entry specifies an initial 12 months, an additional 12-month option and an estimated total value of £151,500. The start date was May 6, 2025.
Those figures are a contract example, not a per-user price. The distinction between awarded and estimated total value matters, too: quoting the larger number as a simple annual fee would misread the record. For a prospective buyer, the sensible comparison includes implementation, training and checking time alongside the software charge.
Beam has also attracted investment backing: Isomer Capital lists it among its portfolio examples. The commercial proposition and the social purpose now share a purchaser in many deployments. Councils want staff capacity and workable records; residents want responsive support. Whether both benefit depends on how the organisation uses the time it gets back.
The person still signs it off
Dundee offers a close view of that arrangement. In an interview published by Beam on August 29, 2026, practice manager Neil Wallace described expansion from Community Justice into Children’s and Adults’ services. The council had made around 20,000 recordings since April 2025. Practitioner feedback averaged 4.26 out of five.
The less headline-friendly detail is the important one: Dundee fitted Beam’s output to its existing templates and Mosaic case management system. Workers read, amend and approve the draft. “We don’t take the human out of the creation,” Wallace said. A product can sit alongside an established system, which makes adoption a different proposition from replacing the whole record-keeping apparatus.
“We don’t take the human out of the creation.”
Neil Wallace / Dundee City Council
Swindon’s published privacy arrangements make the same boundary visible from a resident’s side. Staff explain the recording at the start of a call. If someone objects at that point, the tool is not used. Its children’s services notice explicitly identifies the council’s case management record as authoritative and requires staff to quality-check the draft.
These details define the conditions for useful automation. It needs a suitable conversation, an acceptable recording process, a template that fits and a practitioner with time to check. Skipping that last step would change the service being described. A confident paragraph is still a draft until somebody responsible has read it.
A former caseworker goes to Melbourne
Beam’s expansion has carried some of its frontline experience overseas. Its Australia launch interview introduces Abby, Chloe and Joanna, three longstanding staff moving from London to Melbourne. Joanna had started as a Beam caseworker. The company’s account emphasises spending time with local users and finding the problems that most need attention.
One Australian deployment began in November 2025, when Life Without Barriers piloted Notes with 26 National Immigration Support Service staff. Beam’s case study reports average savings of 4.3 administrative hours per person each week. It also describes hands-on training and troubleshooting. Those unglamorous activities belong in the explanation of the result, rather than outside it.
Beam’s hiring materials describe a culture of ambitious targets, urgency, ownership and improvement. Its rebrand announcement reports almost 100,000 frontline workers across five countries using its technology. That is a company reach claim, not a count of independently verified paying seats. The more useful question for the next deployment remains local: which task, for which team, improves?
There is plenty here to copy without attempting the whole company: start with work you understand, observe the repeated burden, test a narrow intervention and measure the finished task. Keep the people doing it involved when the first design disappoints. Beam’s path began with an electrician’s training. Its next chapter will be judged in ordinary working days, one checked record and one useful conversation at a time.