Violet and pearl rings linked around an apricot sphere

Your tools,
working as one.

We build automations, AI assistants and integrations for small and mid-size businesses. We help your team handle routine work, answer questions and move information between apps.

Let’s talk about your work
See a sample workflow:Support emailsOrdersInquiries
In every sample, a person makes the decision

What we build

We connect the software you already pay for and build the pieces that are missing. Most projects fall into one of these five areas.

Services and sample projects

Automations

Steps your team repeats every day, done the same way every time.

For exampleWhen a booking form comes in, create the task, send the reminder and update your records.

AI projects

AI that reads, sorts or drafts, with a person checking the result.

For examplePull the key details from a long document for someone to check.

Chatbots and assistants

Answer common questions for customers or staff, and pass the hard ones to a person.

For exampleA chat on your website that answers “Where is my order?” by checking your store.

Integrations between your systems

Your tools share the same customer, order and invoice details.

For exampleSend a new order to your customer records and accounting tool.

Customer support automation

Emails arrive sorted, with a draft reply waiting for your team.

For exampleSort an email, find the right help article and draft a reply for your team to approve.

See it working

Three sample workflows. AI prepares the work, your tools move it, and your people make the decisions.

Choose an example workflow
Continue to How we work

A reply, ready for review.

A customer email arrives. AI sorts it and drafts a reply from your approved help articles. Someone on your support team checks and edits it before anything is sent.

If details are missing, the email goes straight to a person.

AI preparesA person decidesSample data
  1. “Can I change the delivery address for order #2381? We moved last week.”

    AI sorted itTopic: Delivery changeUrgency: HighRoute: Support
  2. How the AI got there

    1. Reading the emailOrder #2381Change address
    2. Looking up order #2381Not shipped yet
    3. Searching your help articles

    FoundDelivery changes article

    AI draft

    Draft: “Yes. Order #2381 hasn’t shipped yet, so we can update it. Please reply with the new address.”

    Source: Delivery changes article

  3. Order #2381 is already packed. We can still change the address until 3 pm today.

    AI rewrote it from your note

    Order #2381 hasn’t shipped yet, so we can update it.

    Edited: “It’s packed, but we can still change it. Please send the new address by 3 pm today.”

    Approved

  4. The customer gets the edited reply. A copy is saved with the ticket.

    Sent

    AI filed itSaved to the ticket

    Suggested for your team to approveUpdate the Delivery changes article: packed orders need the new address by 3 pm.

Another example: Orders · Inquiries

Enter an order once.

An order arrives in your store. AI checks it for problems. The customer record updates, and an invoice draft goes to accounting. Your team checks flagged orders before an invoice is issued.

AI checks each new order for things you want to know about, such as a delivery address that differs from billing. Fixed rules move the order, and every update carries the same order number.

AI preparesA person decidesSample data
  1. Order #1042 · 2 items · Paid

  2. How the AI got there

    1. Comparing with the customer recordAddress differs from billing
    2. Checking the quantity2 items: normal

    FoundCustomer record for order #1042

    AI check

    Flag: the delivery address differs from billing. Quantity is normal. Your team checks this order before the invoice is issued.

    Source: Customer record for order #1042

    Normal orders? No team check needed.

  3. Order #1042 is added to the existing customer record.

  4. Draft INV-0193 · Order #1042 · 2 items

  5. Flag: delivery address differs from billing. Check before issuing.

    Checked and issued

  6. Invoice INV-0193 · Order #1042 · Sent to customer

    Issued

Another example: Support emails · Inquiries

Each inquiry goes to the right person.

A new inquiry comes in through your website form. AI reads it, pulls out the facts and suggests who should handle it, based on rules you set. Someone on your team confirms and takes it from there.

If a request is unclear, it goes to your team instead of getting lost.

AI preparesA person decidesSample data
  1. “Hi, we run two cafés and need about 400 printed cups a week from next month. Can you send prices? Do you deliver on Saturdays?”

  2. How the AI got there

    1. Reading the messageQuote request400 cups a weekSaturday delivery?
    2. Matching your routing rulesOwner: Sales team

    FoundYour rule: quote requests go to Sales

    AI summary

    Summary: quote request, about 400 cups a week, starts next month, asks about Saturday delivery. Suggested owner: Sales team.

    Source: Your rule: quote requests go to Sales

    Unclear request? Your team inbox

  3. Sales is the right owner. Add the Saturday delivery question to the follow-up.

    Owner confirmed

  4. Follow-up task for Sales, with the original request and the Saturday delivery question.

    Assigned

    AI addedAI summary attached

Another example: Support emails · Orders

Tell us about a workflow like this

Tools we connect

You keep your tools. We connect them in the ways they already support.

We can connect common tools such as Gmail, Outlook, Slack, Microsoft Teams, HubSpot, Shopify, QuickBooks, Xero and Google Sheets. We check how each one connects before we suggest a build. These names are only examples. We are not partners of these companies, and they do not endorse us.

  • Help desktoChat

    An urgent ticket posts to the team channel with a link.

  • CalendartoCustomer records

    A booked call adds a note to the customer record.

  • Shared inboxtoHelp desk

    An email with an order number opens a ticket with the order attached.

  • EmailtoSpreadsheet

    Details from an attachment become a new row.

  • Customer recordstoEmail

    When a follow-up is due, the assigned teammate gets an email reminder.

Most business software can share information in some way. We check how yours does before we suggest a build.

How we work

Start small. Make it useful.

  1. Map the work

    We follow one process from start to finish with the people who do it.

    You get A one-page map of the process, with the part worth changing first.

  2. Build the first piece

    We build a working version with your tools and try it on real examples, including the awkward ones.

    You get A first version, tested on your own examples.

  3. Launch with people in charge

    Your team tries it first. Then we switch it on, with clear rules for exceptions.

    You get A written list of what runs on its own and what needs approval.

  4. Keep improving

    After launch, we check how it works in daily use and fix what gets in the way. You choose how much support you want after that.

    You get Improvements based on everyday use, with clear instructions for your team.

Ways to work with us

You don’t need a detailed plan to get started. We can help you choose a good first project, or fix something you already use.

Compare ways to work with us

Review

Not sure where to start?

We look at one process with you, check your tools and the risks, and give you a clear list of next steps.

A one-page map of one process, with clear next steps.

Ask about a review

Build

Have a workflow in mind?

We build and test one automation, assistant or integration with the people who will use it.

One workflow, from mapping and testing to launch and improvement.

Ask about a build

Improve

Already have something running?

We look at an automation or process that is unreliable, slow or hard to maintain, then fix the parts that matter most.

We start by mapping the automation or process you already use.

Ask about an improvement

Questions people ask

Short answers before a first call.

Ask us in the form
How do you handle our data?

We agree up front which data the workflow needs, who can see it and which services process it. We ask only for the access the job needs. Before we build, we also agree how long data is kept. Please don’t include sensitive records in your first message.

Can you work with our tools?

Tell us what you use and we will check how each tool can connect. If a connection is not practical, we will tell you before you commit to anything.

How long does a first project take?

We keep first projects small on purpose: one process, one useful change. After we understand the process, we give you a timeline in writing. The main things that affect it are how many tools are involved and how quickly we can get access.

What does it cost?

We give you a written quote after we understand the process. The price depends on how many tools are involved, how many rules the process has and how much support you want afterward. Fees for the tools themselves, such as AI usage, are separate, and we list them up front.

Who owns what you build?

We agree this in writing before we start. Our aim is a tool your business can run, with instructions and the access your team needs. Any third-party subscriptions and licenses are identified up front.

What if the AI is unsure?

Then it hands the work to a person. Before launch, you decide what the AI may do on its own, such as answering “Where is my order?”, and what must wait for a person to approve. We test missing details, unusual requests and mistakes, not only the easy cases.

Ask us in the form

What could
work better?

Tell us about the task your team repeats by hand, the tools that don’t share information, or the idea you want to try.

Please don’t include passwords, payment details or customer records.

Your message comes to us by email. We reply to the address you give us. Read our privacy notice.