AI that works inside your systems, not next to them
Artificial intelligence only pays off when it sees real data and is allowed to do something useful with it — without anyone losing control. So we build our AI agents on top of business systems, connect them through APIs, let them act as independently as you decide, and leave a trail for every action. This page describes that approach and shows where AI works today in each of our products.
- Agents on ERP, CRM and sales
- Autonomy you choose
- Human approval where risk is
- A trail of every decision
Six rules we design AI by
The rules come from the products we already offer — the sales agent, the voice agent and the assistant in NG CRM — and they guide how we are developing NG Tara.
Inside the systems, through APIs
The agent reads and writes through the same API as the rest of your applications. No parallel database, no re-keying, no spreadsheets between steps.
Autonomy is chosen, not assumed
An agent can only suggest a move, prepare it and wait for approval, or act on its own within limits you set — and the level can differ by process.
A person confirms what carries risk
A write, a send or a change that carries risk goes through confirmation until the company decides otherwise. In NG CRM, sensitive actions also need a second person, and whoever proposes never approves their own request.
A trail of every decision
What the agent saw, what it proposed, who approved it and what was actually written — all of it stays in a trail you can still check a year later.
Your data, your keys
An assistant can do no more than the user it acts for. Where the product supports it, as NG CRM does, the company picks its own AI provider and enters its own key.
Measure first, then expand
Every process gets a baseline before rollout. Autonomy grows only once the numbers, and the people working with the agent, confirm it is justified.
Three layers, one logic
AI agents work on top of business platforms, and services connect both to the systems a company already has.
01 · Agents
AI agents that run the work
NG Sara runs the sales conversation and looks for opportunities, NG Nora takes incoming calls, and NG Tara runs business workflows across several systems and channels.
02 · Platforms
AI built into the platforms
NG Copilot in NG CRM answers from your data and writes only after confirmation. NG One builds AI for questions, forecasting and document processing into its core, and NG Commerce is where Sara and Nora operate.
03 · Services
AI shaped to your processes
When a ready-made product does not cover a process, we design an AI solution or automation on your data — with accuracy evaluation and human control at the sensitive steps.
From the first suggestion to working on its own
Autonomy is not switched on on day one. Every step ends in production and opens the next one only once the previous one has proven itself.
- 01
Map the process and the data
We establish where a decision is made, what data it needs and which system holds that data. Without this the agent has nothing to work on.
How data flows across systems and teamsDecision points and who makes them todayBaselines: time, errors, volume - 02
Copilot: the agent suggests
The agent prepares a reply, an offer or an action, and a person decides. The team sees how the agent reasons before trusting it with a single move.
Suggestions with their data sourceThe team rates the suggestionsLimits: prohibited topics and actions - 03
Approval: the agent prepares, a person releases
Routine steps move faster, but each still passes a confirmation. Thresholds for amount, discount or risk decide who approves.
One-click confirmationA second person for sensitive actionsA trail of who approved what - 04
On its own, within limits
Where the numbers have confirmed it, the agent works alone — with limits, oversight, and the option to step autonomy back at any moment.
Limits per action, amount and timeOversight and regular trail reviewsSwitch-off without stopping the business
Where AI works today
An overview of AI capabilities by product — each one is part of what we contract today.
| Capability | Status | Note |
|---|---|---|
| NG Sara — autonomous sales agent | On offer | Three autonomy levels and a full audit trail; part of the NG Commerce offering. |
| NG Nora — AI voice agent | On offer | Incoming calls, hand-off to a live agent with context, transcript and audit trail. |
| NG Copilot in NG CRM | On offer | Answers with a source, writes only after confirmation; the company picks and enters the AI key. |
| Bespoke AI solutions and automation | On offer | A service: agents, RAG, document processing, with evaluation and human control. |
| NG Tara — business workflow agent | On offer | Voice, chat, email and messaging; work through APIs and MCP in ERP, CRM, WMS, DMS and MES systems, with human approval. |
| AI in the NG One core | On offer | NG Copilot for business questions, forecasting, deviations, OCR and an agent for routine steps. |
| NG Copilot and MCP in NG DMS and NG WMS | On offer | Answers citing the exact page in documents, questions on stock and orders in the warehouse; an MCP server for your team's AI assistants. |
“On offer” means the capability is part of a product or service we contract today. The first step is a demo on your own process.
Writing on AI agents and automation
- AI & automation5 min read
NG Tara: an AI agent for business workflows — how it runs a task across several systems
NG Tara is an autonomous AI agent for business workflow automation. What it does, through which channels and with which systems, under which rules — and the value it brings to a team.
Read - AI & automation7 min read
An AI strategy for the enterprise: five decisions before the first model
An AI strategy is not a list of tools but five decisions: where AI reads data, what it may do, who confirms, who holds the keys and in what order it is rolled out. How we make them, and where our products sit in them.
Read - Security10 min read
Security and privacy of AI agents: what to check before production
An AI agent in production has access to data and takes actions — which makes it a new attack surface. A checklist: permissions, prompt injection, tool guardrails and the fate of personal data.
Read
AI strategy — FAQ
What does an AI strategy mean for Konis Software?
An approach in which AI agents work inside the business systems a company already runs — ERP, CRM, sales and communication — through APIs, with autonomy chosen by the company, human approval for risky actions and a trail of every decision. It is introduced step by step: from an agent that only suggests to working on its own within set limits.
Can an AI agent write data or send messages on its own?
Only as far as you allow. NG Sara has three autonomy levels — suggestion, approval, or acting alone within limits — and the level is chosen per segment. NG Copilot in NG CRM writes nothing without the user's confirmation, and sensitive actions also need a second person.
Which AI products are on offer today?
NG Sara (sales agent) and NG Nora (voice agent) as part of the NG Commerce offering, NG Tara for business workflows, NG Copilot in NG One, NG CRM, NG DMS and NG WMS, MCP servers for the AI assistants your team already uses, and bespoke AI solutions and automation as a service.
What is NG Tara?
NG Tara is an autonomous AI agent for business workflow automation: communication over voice, chat, email, Viber, WhatsApp, Instagram and other channels, and work through APIs with ERP, CRM, WMS, DMS and MES systems. Which actions the agent may take alone and which need confirmation is set by you, per process.
Where does our data stay, and who picks the AI model?
In NG CRM the assistant works within the permissions of the user it acts for and can do no more than they can; the company picks its own AI provider and enters its own key, and company rules decide which operations and which data may reach the model. For bespoke solutions we agree the same framework at the start of the project.
How do we know AI pays off before we expand it?
Before rollout, the process gets a baseline — time, error count, volume of work. The agent then works at suggestion level, then with approval, and autonomy grows only once the numbers and the team confirm the value. Results are not promised in advance; they are measured on your process.
Let's start with one process
Describe the process that costs you the most. In a short call we tell you where an AI agent would make sense, at which autonomy level it would start, and what would be measured.