“Anyone in?” Now you’ll know before the van leaves.

AnyoneIn is the AI agent that checks your customer will be in for any delivery that needs them: signed-for, age-checked, two-person, installs and collections. It asks by email, then text, then a natural phone call, and rebooks anyone who can’t make it.

Tonight on this street: 9 confirmed2 rebooked0 wasted trips

Illustrative

Illustration: a street at dusk. As each customer confirms, their window lights up amber, and the van only stops at the lit houses.

The knock nobody answers.

One missed knock costs the trip, the phone calls, the apology, a second trip, and sometimes the order.

For parcels, a failed delivery costs about £11.60, and a lost order about £147.1 Deliveries that need someone in cost far more.

What one failed visit costs

Furniture deliveryOur default assumptions

  1. Wasted crew and van slot£380 crew day ÷ 12 drops£31.67
  2. Redelivery slot and handlingrest of the £55 per failed visit£23.33
  3. Customer-service contacts2.5 × £4.50£11.25
  4. Goodwillthe apology voucher£10.00
  5. Cancelled-order margin5% × £900 × 35%£15.75

Total≈£92

IllustrativePut in your own numbers

Collections have it worse.

The customer isn’t home, or the item isn’t ready or boxed. The refund is delayed, the customer chases, and the returned stock loses value while it waits.

A failed delivery

One wasted trip, and a second one booked

  • Nobody in to sign
  • Crew slot wasted, redelivery booked
  • Complaint calls, goodwill

AnyoneIn confirms before the van leaves, and rebooks anyone who can’t make it.

How deliveries work

A failed collection

Everything stalls until someone sorts it

  • Customer not in
  • Item not boxed or not ready
  • Refund delayed, so the customer chases
  • Returned stock losing value while it waits

AnyoneIn checks someone’s home and asks for a photo of the item ready to go, for retailers and the carriers who collect.

How collections work

Today the fix is someone ringing round the day before.

Most teams handle this by calling customers the day before. It takes hours, and plenty of people don’t pick up calls from numbers they don’t recognise.

Illustrative Calling 40% of 4,800 jobs a month at 4 minutes each is 128 staff hours, before the voicemails and the call-backs.

And volumes keep growing: two-person delivery alone is up about 6.5% a year, against about 2% for wider retail.2

4:00pmthe day before
Calls waiting7
Tomorrow’s drops confirmed38 of 146

It starts with an email and only rings if it has to.

If the email gets no reply it sends a text, and if that goes unanswered it phones, always in sensible hours and on your brand. The call reaches the people who ignore texts, and it can rebook them there and then.

  1. Three days before

    Email

    A branded email with three one-tap buttons: I’ll be in, someone else will be in, I need a different day.

    Typical reply: Taps “Yes, I’ll be in to sign”

  2. Two days before, if there’s no reply

    Text

    A short text. They reply in their own words, and the agent understands.

    Typical reply: “At work but my partner Jo will sign”

  3. The day before, if there’s still nothing

    AI call The step others skip

    A natural phone call, in a British voice, that reaches the people who ignore texts. It confirms, rebooks or takes a name, and leaves a voicemail if a machine answers.

    Typical reply: “Tuesday’s no good. Could you do Thursday?”

  4. As soon as there’s an answer

    Outcome, back to you

    A clear outcome, the reason behind it and the next action, in a signed webhook to your systems. The job’s full timeline is there through the API.

Any reply pauses the plan. A clear answer closes it.

Hear it rebook a delivery in under a minute.

Sam can’t do Tuesday, so the agent offers the slots you allow and books Thursday morning. It checks it’s speaking to the right person first, and only mentions the first line of the address.

  • Says it’s automated, up front and when asked
  • Confirms identity before sharing any details
  • A British voice, in plain British English
  • Restates the outcome, then lets them go
Harlow & Finch assistant (AI) calls SamScripted demo call · Harlow & Finch is a fictional retailer
  1. Harlow & Finch assistant (AI)

    Hello, this is the AI delivery assistant for Harlow & Finch, calling about your delivery. Am I speaking with Sam?

  2. Customer

    Yes, that’s me.

  3. Harlow & Finch assistant (AI)

    Thanks, Sam. Your Aldgate three-seat sofa is booked for Tuesday, between 8am and midday. Will you be in to sign for it?

  4. Customer

    Ah, no. Tuesday’s no good, I’m away with work. Sorry, am I talking to a real person?

  5. Harlow & Finch assistant (AI)

    I’m an AI assistant, but I can sort this for you. I can do Thursday morning, 8 till 12, or Friday afternoon, 12 till 5.

  6. Customer

    Thursday morning would be perfect.

  7. Harlow & Finch assistant (AI)

    Lovely. That’s Thursday the 15th, between 8am and midday. Someone over 18 needs to be in to sign for it.

  8. Customer

    That’s fine. Same address, yeah?

  9. Harlow & Finch assistant (AI)

    Yes, the address on your order, on Elm Road. You’ll get a text confirming the new slot in a minute.

  10. Customer

    Brilliant, thanks.

  11. Harlow & Finch assistant (AI)

    Thanks, Sam. Thursday morning it is. Goodbye.

RebookedTue 13 Oct, 8–12 → Thu 15 Oct, 8–12 · by AI call

Customers reply however they like, and it still gets it right.

Someone else signing, Thursday instead of Tuesday, a question, a complaint or a STOP: the agent works out what each reply means, notes what your crew needs and takes the next step.

  1. Confirmed1

    “Yes, I’ll be in all morning”

    Will be in

  2. Someone else signing1

    “My flatmate Priya can sign for it”

    Signing: Priya

  3. Rebooked1

    “Not Tuesday, sorry. Thursday morning?”

    Moved to Thu 8–12

  4. Needs a person1

    “The last one arrived damaged, can someone call me?”

    Next: call the customer back

  5. Opted out1

    “Please stop texting me”

    No more messages, on any channel

It gets the crew ready, and helps at the door.

A yes isn’t always enough. The sofa might not fit up the stairs, the wardrobe for collection might still be standing, or the customer might be five minutes away when the driver knocks. AnyoneIn asks about these before the day, and helps the driver save the visit on it.

The way in, and a photo

After they confirm, the customer’s page asks about the floor, the lift, the stairs, the narrowest doorway and parking. For a collection it asks for a photo of the item ready to go, and an AI check tells your team if something looks wrong.

Will it fit?

Send the item sizes and it compares them with the narrowest point the customer gives, and notes turning stairs, no lift or a parking permit. If it’s unlikely to fit, your team is told straight away.

Help at the door

If nobody answers, the driver taps one button on the job’s driver page. The agent rings the customer at once and agrees the quickest option the job allows, such as a few minutes’ wait or a neighbour who signs.

A record of what happened

Every outcome records why the customer couldn’t make it. For a dispute, one API call returns the whole job: messages, the call transcript, the answers and each photo with a hash that shows it hasn’t changed.

How crew checks and doorstep rescue work

What a month of confirmed deliveries gives back.

Here’s what stopping half of the avoidable failures could mean for a furniture retailer with 4,000 deliveries and 800 collections a month.

IllustrativeOur default assumptions for a furniture retailer. Change any of them in the calculator.

122

failed visits avoided a month

194

customer-service hours back a month

10

crew days freed a month

Try your own numbers

4,800
7%
£55
Illustrative

About this much a month back

≈£7,235

From about 109 failed visits avoided, counting the crew and van, and the customer-service contacts each failure causes.

See the full breakdown

Once you’re live, a monthly report shows the real figures: jobs sorted without a person, why customers couldn’t make it, which channel got the answer, doorstep rescues that saved the visit, failed visits prevented, and an estimate of the CO2e avoided.

Coming soon

The best moment to offer an add-on.

They’ve just confirmed, and they’re thinking about the delivery. That’s when taking the old sofa away, assembly or a one-hour slot is an easy yes.

Only offered where your customers have agreed to hear from you.

How upsells will work
11:04
H&FHarlow & FinchText message

Tuesday’s no good. Thursday morning?

All set for Thursday, 8–12. Want us to take your old sofa away for £25?

Take away your old sofa£25 · added to your order

Done, added to your order.

Plugs into what you run.

Keep your OMS, your carrier and your routing. AnyoneIn adds the one job they don’t do, which is finding out who’s in.

  • KeepYour OMS
  • KeepYour carrier
  • KeepYour routing
  • AddAnyoneIn

Send each order through the API and get the outcome back in a signed webhook. It works with any carrier, because it confirms before the job is handed over, and nothing in your stack needs replacing.

  • APIOne call per orderAvailable
  • WebhooksSigned confirmation.updatedAvailable
  • Your OMS and carrierPrebuilt connectorsPlanned

AvailablePlanned

It speaks for your brand and always says it’s an AI.

The rules it follows on every message and call.

Honest AI disclosure

Every call says it’s your automated assistant. Ask if it’s a person and it says no.

Built in the UK, stored in London

Your data is stored on Google Cloud in London, and the service runs on Google Cloud in Belgium. We process it on your behalf.

Quiet hours and opt-outs

Texts and calls only between 9am and 8pm, with STOP honoured on every channel. Built for UK rules: PECR and Ofcom’s guidance on calls.

A human when it matters

Complaints, damage and “can I speak to someone?” are flagged to your team in the webhook.

Security and compliance

Prove it on your own numbers.

We’re taking on a handful of UK retailers and carriers for a 60-day pilot, with a hold-out group so you can see the difference in your own numbers.

You get

  • A 60-day pilot with your first 2,000 jobs free
  • A hold-out group, so the result is provable
  • Set-up help and a weekly results report
  • A say in the roadmap, and founder pricing for 12 months, confirmed in writing when we sign

We ask

  • A weekly 20-minute check-in
  • An anonymised case study, if it works

Questions retailers ask us.

Anything else? Ask us

Will customers know it’s an AI?

Yes. Every call opens by saying it’s your automated assistant, and if a customer asks whether they’re talking to a person, it says it’s an AI and carries on helping. Emails and texts carry your brand’s name and tone, and a person can always be reached.

What if a customer wants to speak to a person?

The agent tells them your team will call back, marks the job “wants a person”, and sends that, with the next action, to your systems in a signed webhook. Complaints, damage and refunds always go to a person.

Which carriers does it work with?

Any. AnyoneIn confirms with the customer before the job is handed to a crew, so it works from your order data through the API and webhooks, whoever delivers. Prebuilt connectors for delivery platforms are planned.

What does it cost?

You pay per job, however many messages and calls it takes, and only for jobs where we get an answer. Prices are indicative and shown up front, and they depend on the job: £0.95 to £1.40 for a two-person, install or collection job, and £0.35 to £0.55 for a signed-for or age-checked delivery. The 60-day pilot includes your first 2,000 jobs free. See pricing

How long does it take to set up?

The integration is one API call per order, and one webhook endpoint to receive the results. During a pilot we set up your brand, tone and the slots it can offer with you. Prebuilt connectors for order systems such as Shopify are planned.

What data do you need?

Only what it takes to confirm: the customer’s name, mobile and email, the items, the date and window, the address, and the other slots it may offer. Item sizes are optional, for the fit check. On their page, customers can add details about the way in and up to three photos. It’s stored on Google Cloud in London, the service runs on Google Cloud in Belgium, and we act as your processor.

Sources and assumptions

  1. ParcelBroker, “The true cost of failed delivery”. UK parcel figures: about £11.60 per failed delivery and about £147 per lost order. ↩
  2. Meteor Space, “Two-man delivery market statistics”. Two-person delivery growing about 6.5% a year, against about 2% for wider retail. ↩
  3. Figures marked Illustrative use our default assumptions for a furniture retailer (for example £55 per failed visit, a 7% failed-delivery rate and £4.50 per customer-service contact). They’re not measured results. Change any of them in the ROI calculator.

Anyone in?
Find out before the van leaves.

Take a 60-day pilot with a hold-out group, and see the difference in your own numbers.