■ SAFE RISK ■ Personal Services
No — and the demand curve is pointing hard the other way. Accompanying a stranger through the last weeks of their life is presence work, and presence is the one commodity a model cannot ship.
“Guiding someone through dying is the most human job imaginable. AI need not apply.”
Our AI replacement risk score — how we score jobs
The actual work is less mystical than the title suggests. A death doula meets a client after a terminal diagnosis and helps build a plan: advance directives, who is allowed in the room, what music plays, whether the dog can be on the bed. They sit vigil for hours, sometimes overnight, so an exhausted spouse can sleep. They coach families on what agonal breathing sounds like so nobody panics at three in the morning. They handle the unglamorous logistics too — funeral home comparisons, home-funeral rules that vary by state, legacy letters, and the fraught negotiation between siblings who disagree about hospice.
Some of that periphery is already being absorbed by software. Advance-directive platforms generate legally sound documents, estate and funeral-cost comparison tools have been online for years, and chatbots offer grief scripts at three in the morning when no human is awake. Scheduling, care-team coordination and the endless paperwork of dying are ripe for automation and largely welcome it. A model can also brief a new doula on the cultural expectations of a family it has never met.
The core does not budge. Sitting with terror rather than fixing it, noticing that a dying man keeps mentioning his brother in Fresno and quietly making a phone call happen, being physically present when a body stops — these are acts whose value is precisely that another person chose to be there. A synthesised voice offering comfort at a deathbed reads as an insult, not a service. Our score of 8 covers the admin, not the vigil.
Automatability: our editorial assessment of current and near-term AI capability
Resilient indefinitely. If anything the profession is in a growth phase: ageing populations in wealthy countries, overstretched hospice systems, and a cultural shift toward planning death rather than avoiding it are all pushing demand up through the 2030s. Automation will quietly eat the paperwork and the scheduling, which most doulas will regard as a gift. The vigil itself is not a market software can enter.
Viable, though rarely lucrative on its own. Demand is rising with ageing populations and strained hospice capacity, and the role has no automation threat worth naming. The real constraints are payment — most work is private-pay rather than insurance-reimbursed — and emotional sustainability. Many practitioners combine doula work with chaplaincy, social work, hospice volunteering or grief counselling to build a full income.
At the margins. Chatbots can answer questions at hours when no human is available, explain what hospice medications do, or prompt someone through a legacy letter. Some isolated people genuinely value that. But the thing families remember is that a person stayed. Machine presence is not presence, and most hospice teams treat it as a supplement to human accompaniment rather than a substitute.
The administrative scaffolding. Advance-directive generation, funeral-cost comparison, care-team scheduling, insurance and benefits navigation, transcription of memoirs, and reminders for medication or ritual timings. That is real relief — those tasks eat hours that doulas would rather spend at a bedside. Nothing in the near-term stack touches vigil, family mediation or the physical care of a dying body.
Take a recognised training programme, then volunteer with a hospice to accumulate real bedside hours; nothing substitutes for having sat with actual dying people. Learn your state's rules on home funerals, body transport and death certificates. Build relationships with palliative nurses and funeral directors, who supply most referrals. And arrange your own supervision before you need it, not after.