Opportunities › Persistent AI agents for business work automation › Proposed app
Deskhands
Persistent AI agents for scheduling, research and documents
- Score
- 42
- confidence 38
- Evidence
- 4
- items on the page
- Rivals
- 0
- named in sources
- First version
- 45 d
- estimate
- Type
- Desktop and mobile
- PWA: on computers and phones alike
- Market since
- 2023
- Emerging market · AI estimate
ScreensConcept
Desktop and mobile: Agents are set up and monitored on desktop for complex workflows, but approvals and activity checks happen on mobile while knowledge workers move between meetings and offices. Sample screens on a computer and a phone, in the proposed brand. Their numbers are sample data.
Your agents
All within limits3 working
About this appAI proposal
Deskhands gives each routine job its own persistent AI agent. An agent keeps a name, a mailbox, a workspace of files and a log, and it carries on between sessions instead of starting fresh with every request. It can schedule meetings, prepare documents, research the web and run long tasks.
You watch and steer from a simple dashboard: what each agent is doing now, what it finished, and what it is waiting on you to approve. Anything that sends, spends or shares needs a yes from you until you decide otherwise.
It is for small business teams and knowledge workers who want routine work off their plate without wiring together many separate automations. Agents are visible, limited and stoppable.
- Tagline
- Colleagues that keep working after you log off.
- Voice
- CalmPlainDependable
- Why this name
- A desk is where work lives. Deskhands are agents that stay at your desk, keep their own place and can be checked on at any time.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Persistent agents
Each agent keeps its own memory, files, inbox and task history, and picks up where it stopped, including after hours or days.
No member yet
Live activity view
See every agent's current step, recent actions and waiting items in one timeline, updated as they work.
No member yet
Approval gates
Set which actions need your yes: sending email, booking, sharing files or spending money. Approve from your phone.
No member yet
Scheduling
An agent reads your calendar rules, proposes times to guests and books the meeting without back and forth.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#0F3D3EDeep Teal · primary#F6F3ECPaper · surface#1B1F23Ink · text#E8A33DSignal Amber · accent#4F8A5BMoss · data / success#B54A3ABrick · alert
Typography
Aa
Deskhands
PricingAI proposal
The material gives no competitor prices. These tiers are a starting point to test, based on the agent count and compute hours named in the opportunity's monetization idea.
Trial
Free
- 1 agent
- Limited monthly run time
- Approval on every action
Team
$49 a month
- 5 agents
- Calendar, email and file connections
- Activity log and approvals
- Fair monthly compute included
Business
$199 a month
- 25 agents
- Long-running jobs
- Role-based permissions
- Extra compute at usage rates
Would you pay Team · $49 a month?
Your answer tells PWAS which prices people accept. It changes no score.
No answers yet.
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- Shared outputs. Agents send meeting invites, briefs and documents to outside people. Each carries a small 'prepared by a Deskhand' line and a link to the shared file or booking page.
- Do they become customers?
- Recipients are mostly other knowledge workers who could run their own agents, but many will only read a document once. Test how many recipients click through and start a trial.
- First customers
- Founders and office managers at 5 to 20 person service businesses, such as agencies and consultancies, who lose hours each week to scheduling and meeting follow-up.
- Where to reach them
- Hacker News and Show HN: Post a demo of the live agent view; the source thread already shows interest in watching agents work.
- Founder communities: Offer a free setup call in small-business and agency groups, and build one agent per customer by hand.
- Direct outreach: Message agency owners with a specific offer: meeting scheduling and follow-up handled for two weeks.
- First milestone
- Five paying small agencies keep at least one agent running for four weeks.
Why people already pay for this
- VerifiedHacker News post received 21 points and 8 comments
- Public claimAI agents can handle knowledge work tasks including scheduling, document preparation, research, meetings, and long-running business operations
- Public claimExisting agent infrastructure treats agents as API calls rather than persistent processes with dedicated resources
- Public claimThere is demand for observing and managing autonomous agents doing daily work
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Agent runtimePer-agent isolated containers or sandboxesEach agent needs its own persistent storage, compute and network identity, which serverless functions do not provide.
- This app · Job queueDurable workflow and queue serviceLong-running tasks must survive restarts and resume at checkpoints.
RisksAI inference
- AIAgents may take wrong actions, such as a mistaken booking or an email to the wrong person. Approval gates and logs reduce this but add friction that could weaken the value.
- CommercialEvidence is a single Hacker News post with modest engagement and public claims only. Willingness to pay for persistent agents is unproven.
- TechnicalRunning long-lived agents with isolated storage and network identity costs money per agent. Compute may erode margins at low prices.
- PlatformCalendar, email and document connections depend on third-party APIs and permission reviews that can change or slow launch.
- DataAgents hold business emails and files, so a breach or a misconfigured permission would be serious and may trigger customer security reviews.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.