Skip to content
⁨NDDev AI⁩, home page
Community
Language: English
Display
Theme
Motion

Business process automation with AI

NDDev automates business processes with AI for sales, procurement and back-office teams: sorting incoming emails and requests, checking documents and moving data between CRM, accounting systems and spreadsheets. Rules and calculations run as ordinary code, and a language model handles the steps that need to understand text. We built our first AI automations in 2023.

Who it is for

Companies where the same stream of emails, invoices and requests is sorted by hand every day, and managers who want to see how many tasks the system closes by itself.

  • Code and a model in one process
  • Exceptions go to a person
  • Retries when integrations fail
  • Manual work measured before and after

Where the model fits in the process

We break the process into steps and choose a tool for each one. Rules, calculations and field transfers are ordinary code: faster, cheaper and covered by tests. The model takes the steps that involve free text: working out what an email is about, pulling details from an invoice, matching a request to a customer, drafting a reply.

  • a process map with the volume and time of manual steps
  • integrations with CRM, accounting, email and messengers
  • classification and routing of incoming messages
  • field extraction from emails, invoices and requests
  • an exceptions queue for your staff
  • a report on the share of tasks closed automatically

Connections to your systems follow the rules of reliable integration: a task queue, idempotent processing, retries with growing delays and a log of every call. When the model is unsure or the data disagree, the task lands in an exceptions queue with a note on what to check. The report shows the share of such tasks, and each resolved case helps us refine the rules and prompts.

Moving the flow to automatic processing

We start by measuring, so the effect of automation shows in numbers.

  1. Measurement

    We count how many tasks arrive, how long each one takes and where mistakes happen most.

  2. Design

    We split the steps between code, the model and people and fix this in the specification.

  3. Integrations

    We connect the systems, set up queues and retries and build a test flow from real examples.

  4. Parallel run

    The automation works alongside your staff while we compare results and adjust the rules.

  5. Switch-over

    The flow moves to automatic processing, and exceptions go to a queue for your staff.

Questions about AI automation

Which processes should we automate first?

Frequent, repetitive ones with a clear outcome: sorting incoming requests, reconciling documents, moving data between systems, standard replies to customers. The more such tasks you have each day, the more visible the result.

Will we have to change our systems?

Usually not. We connect to what already works through APIs, exports or a mailbox. If a system has no API, we discuss a workaround and its risks before work begins.

What if the model gets a document wrong?

Each step has a confidence threshold. Below it, the task goes to a staff member with the doubtful field highlighted. Every action is logged, so any error can be found and reviewed.

How does AI automation differ from ordinary automation?

Ordinary automation works with structured data under fixed rules. A model adds the steps that involve free text: emails, comments, scans, chat threads. In our projects the two work together, and each step goes to the tool that handles it more cheaply and reliably.

Who looks after the automation after launch?

We always arrange support: improvements, updates and a monthly report, usually for 10% of the project price per month. During support we add new scenarios and keep the integrations working as your systems change.

How much does automation cost?

The price block shows where pricing starts. The final figure comes from the number of steps and systems, the volume of the flow and where the model has to run. Model usage is billed at cost. We fix the exact price and timeline in the specification, which is part of the project price.

from ₽280,000

Free warranty
40 days

Prices are starting prices for information only and are not a public offer. The final price is fixed in the specification.

AI adoption for business

Send us an example of a task done by hand

Describe the process or forward an anonymised example of an email, request or document. We will estimate how much of the flow automation can take over.

We reply within 24 hours

Assistant

Loading the assistant…