Learn

Seller cases, product stories, and how-tos — all in one place

All

Post

Release

Case

Press

All topics

Marketplaces

CRM & calls

AI trends

PostSamreshuuuSeptember 22, 2026

AI agent mistakes: what happens when it gets no answer from a marketplace, inventory system or bank

The costliest mistake an AI agent can make is not "didn't do it" but "did it twice". A write through an API has three outcomes: succeeded, refused and unknown, and retrying an unknown outcome creates double payments, orders and receipts. Here is how Samreshuuu handles it with one rule for every integration, how it checks writes in YooKassa, MoySklad and Ozon, and six questions to ask an AI agent vendor.

PostSamreshuuuSeptember 21, 2026
AI for accountants: what it already does in 1C, the bank and e-documents, and what not to hand to it
A chat model and the 1C:ITS consultant answer questions but cannot see your books. An agent with access to 1C, the bank and e-document exchange does the work itself: reconciles the statement with the ledger, finds duplicate payments, builds reconciliation acts and a payment calendar, tracks tax authority demands in Kontur.Extern. We walk through an accountant's month task by task, what not to hand to AI, how to give access to 1C safely, and where to start.
PostSamreshuuuSeptember 17, 2026
AI hallucinations: why the model was wrong yesterday and right today, and how to check an answer before you trust it
A hallucination is a confident answer with no data behind it, and you catch it by asking where the number came from. But an agent that works regularly has a subtler failure: right today, wrong tomorrow, on the same data. By IBM's measurement an agent with 77% accuracy solves only 53% of tasks on every run, and a bigger model does not fix it. We cover three kinds of failure, the flat-decision mechanism, four minute-long checks of an AI answer, and how repeatable work is pinned in our product.
PostSamreshuuuSeptember 15, 2026
How to choose an AI agent for your business without giving it too much: a checklist on data, access and 152-FZ
Even OpenAI calls the mode where its Codex agent does everything without asking dangerous. An agent with access to your accounts carries more risk than a chatbot. Here is what Russia's 152-FZ requires from a company adopting AI, plus a 20-question vendor checklist on data, access, actions and honesty: the good answer and the red flag based on how real agents are built, and a half-hour hands-on test.
PostSamreshuuuSeptember 14, 2026
Don't pay for people who log in once a month: five free seats in an AI agent for your team
The CRO of Databricks says per-user pricing is a thing of the past: once a company pays for every person, it starts deciding who is allowed in, and usage drops. In a small business this is even sharper: an owner won't buy a seat for the accountant who logs in twice a month. Here is how much an AI agent for a team costs at Samreshuuu: a paid seat for whoever works with it daily, up to five free seats for everyone else, and the rule that turns one into the other.
PostSamreshuuuSeptember 12, 2026
An overnight task: handing an AI employee work that runs while you sleep, and waking up to a result rather than a surprise
The author of Bun leaves an agent overnight with one sentence: "I'm going to sleep, in the morning I want to see result X". We translate the trick into a business owner's language: can an AI run tasks inside your Ozon account and 1C without you, how an overnight task differs from a chat and from an AI agent on a schedule, how to name a result that is checked in a minute, where the report arrives in Telegram and what arrives on the way, where the agent must stop on the irreversible and why everything irreversible goes into drafts. With a ready task text for a MoySklad-to-Ozon stock reconciliation.
PostSamreshuuuSeptember 8, 2026
AI in Excel: three kinds of helpers, and why a wrong number does not come from a formula
How an assistant inside Excel differs from a chat with an uploaded file and from an agent that reads the accounting system and marketplace accounts itself; three causes of a wrong number (ambiguity, staleness, a similar column); a three-sheet file that survives the next export; a scheduled task with blanks instead of zeros; the check before the Monday meeting and three scenarios people start with.
PostСам РешуSeptember 7, 2026
How to connect AI to MoySklad: stock, cost prices, and purchase orders
Samreshuuu connects AI to MoySklad with an employee API token: the agent calculates stock, dead stock, margin and shortages by the rule you described in plain words and carries the job through to a draft purchase order. A walkthrough: four ways to connect AI to MoySklad — a connector wrapper, built-in reports, a custom script and an AI agent — how their outcomes differ, and seven places where a MoySklad integration returns a successful response with a wrong number.
PostSamreshuuuSeptember 4, 2026
An AI agent failed halfway through a task: how to resume without doing it twice
What happens when 1C has created an invoice but its response is lost: how checkpoints, heartbeats, idempotency keys, and an audit log help an agent resume after failure, avoid a duplicate write, and stop when the outcome cannot be proved.

AI agent mistakes: what happens when it gets no answer from a marketplace, inventory system or bank

AI for accountants: what it already does in 1C, the bank and e-documents, and what not to hand to it

AI hallucinations: why the model was wrong yesterday and right today, and how to check an answer before you trust it

How to choose an AI agent for your business without giving it too much: a checklist on data, access and 152-FZ

Don't pay for people who log in once a month: five free seats in an AI agent for your team

An overnight task: handing an AI employee work that runs while you sleep, and waking up to a result rather than a surprise

AI in Excel: three kinds of helpers, and why a wrong number does not come from a formula

How to connect AI to MoySklad: stock, cost prices, and purchase orders

An AI agent failed halfway through a task: how to resume without doing it twice

Archive