AI agents designed for measurable workflows

We start from an observable process, define data, checks and exceptions, then validate the agent with KPIs and go/no-go criteria before release. The agent proposes, the person decides.

  1. 01 Process and starting measurement
  2. 02 Checks and exceptions
  3. 03 Pilot and go/no-go decision
Show us a process Explore the products

Our AI products

Agents and tools for marketing, sales, support and data. Features, timelines and service levels of the agents are confirmed on the real scope: the cost is set in a written quote after the analysis, and the first 30-minute call is free. ChatAziendale and Keel have their own pages and terms (in Italian).

Marketing AI Agent

Content and campaigns with human approval

Supports content production, campaign orchestration and performance analysis across channels. The scope is set from authorised sources, approvals, budget, available platforms and KPIs such as production time, cost per asset, quality and attributable conversions.

Key features

  • Content generation: blog posts, social posts, emails, video scripts
  • Multichannel campaigns: Facebook, Instagram, LinkedIn, Google Ads
  • A/B tests: variants proposed and measured on the same audience
  • Performance analysis: trends by channel, content and audience
  • Segmentation: messages adapted to audience segments
KPI
Measured before the pilot
HITL
Human approvals
LOG
Traceable actions

Cost: written quote after the process analysis.

Request a demo About the agent

Sales AI Agent

Qualification, follow-ups and an up-to-date CRM

Supports lead qualification, follow-ups and CRM updates according to agreed rules. The pilot measures data coverage, routing quality, admin time and the rate of human intervention; messages, consent and sending thresholds stay governed.

Key features

  • Lead scoring: contacts assessed against agreed criteria
  • Follow-ups: personalised email and call sequences
  • CRM integration: Salesforce, HubSpot, Pipedrive
  • Suggestions for the salesperson: next steps proposed, the decision stays with the person
CRM
Operational source
SLA
Measured times
QA
Checked sample

Cost: written quote after the process analysis.

Request a demo About the agent

Voice AI Agent

Phone requests with escalation to an operator

Handles well-defined phone requests, gathers information and routes cases by intent, priority and escalation rules. Accuracy, languages, phone system compatibility and hours covered depend on the configuration and are tested on representative calls before go-live.

Key features

  • Natural conversation: understanding of context and intent
  • Several languages: languages and accents checked on test calls
  • Tone detection: escalation to an operator when needed
  • Routing: the call is transferred to the right person
  • Phone integration: compatibility with your PBX checked before go-live
INT
Intents tested
ESC
Human escalation
QA
Sampled calls

Cost: written quote after the process analysis.

Request a demo About the agent

ChatAziendale

An AI chatbot on your company documents

A no-code chatbot connected to company documents and web content, with source citations and custom branding. The project defines authorised sources, content updates, answers that are not allowed, privacy, escalation and a sample of questions to check quality before publishing.

Key features

  • On your company documents: PDF, DOCX, TXT or your website
  • Source citations: answers can link to the documents used, so readers can check them
  • Branding: your colours, logo and tone of voice
  • One-line install: a snippet to paste into your site, no developer needed
  • Several languages: detects the user's language and answers in more than 50 languages
  • Privacy and security: sources, access, retention and data handling configured per project
5 min
Setup
50+
Languages supported
Three monthly plans Free, €29, €149

No credit card needed for the Free plan. Plans and terms on chataziendale.it (in Italian).

About ChatAziendale Try the Free plan (opens in a new tab)

Keel

An MCP layer for Amazon Seller and Vendor

Keel connects Amazon Seller Central, Vendor Central, SP-API, Ads API and Brand Analytics to AI agents such as Claude and ChatGPT through the Model Context Protocol (MCP). You can query Amazon data in plain language, combining accounts, brands and marketplaces in one conversation. It includes seasonal forecasting, exports (Excel, CSV, Google Sheets, PDF, Notion), multi-currency reconciliation, per-agent permissions and a log of every query. It is built for sellers, vendors, multi-account agencies and e-commerce teams who want to stop exporting CSV files by hand.

Key features

  • Multiple accounts and marketplaces: different brands and marketplaces in one conversation
  • Structured access: sales, catalogue, orders, advertising and finance exposed to AI agents
  • Cross-account Brand Analytics: queries that combine data from different accounts
  • Seasonal forecasting: demand and revenue with seasonal models
  • Exports: Excel, CSV, Google Sheets, PDF and Notion straight from the results
  • Permissions and logs: per-agent permissions and a record of every query
8+
Amazon sources unified
1-2 days
Initial setup

Cost: quoted after a demo on your accounts.

About Keel Book a demo (opens in a new tab)

Custom AI Agents

When no standard solution covers the process

We design custom AI agents when the process is not covered by a standard solution. Before building, we check inputs, outputs, integrations, permissions, exceptions and owners; architecture and service levels depend on the approved scope.

Example industries

  • Healthcare and pharma: data permissions and sector rules
  • Finance and insurance: risk checks and approval steps
  • Retail and e-commerce: catalogue, orders and customer requests
  • Manufacturing: production, quality and supply chain
  • Public administration: requests and documents from citizens
INPUT
Sources and permissions
OUTPUT
Expected quality
OWNER
Operational control

Cost: written quote after the process analysis.

Show us a process About our consulting

4 checks to validate an AI agent on the real process

Value is not assumed from the demo. It is measured on a representative sample, including errors, exceptions and human intervention.

  1. 01

    Process and starting measurement

    Volume, time, cost, errors and owner of the workflow, measured before automation.

  2. 02

    Acceptable output

    Quality criteria, authorised sources and the cases where the agent must stop.

  3. 03

    Checks and exceptions

    Permissions, logs, thresholds, escalation to a person and incident handling.

  4. 04

    Go/no-go decision

    KPIs compared with the starting measurement, an owner, a review schedule and an exit plan.

What the pilot delivers: evidence, limits, running costs and a reasoned decision on what goes into production.

Show us a process

Frequently asked questions

Timelines, requirements, customisation, support, security and cost of the AI agents.

How long does implementation take?

Timelines depend on the process, data, integrations, permissions, checks and tests. After the analysis we set a roadmap with milestones and go/no-go criteria, without promising a standard duration before we know the scope.

  • Analysis: process, starting measurement, data, integrations, owners and constraints.
  • Pilot: representative sample, quality criteria, errors and exceptions.
  • Go-live: only after the agreed checks are passed.
  • Operations: owner, monitoring, escalation and review schedule are documented.

How much does an AI agent cost?

The first 30-minute call is free. The cost of an agent depends on the process, volumes, integrations and service level, and is set in a written quote after the analysis. ChatAziendale, instead, has monthly plans published on its own site.

What are the minimum technical requirements?

Requirements change with the use case and the systems involved. We check them before the pilot, without assuming that every solution is cloud-only or needs no change to the existing architecture.

  • Authorised, up-to-date data sources with an identified owner
  • APIs, connectors or import procedures that are available and testable
  • Identities, roles, permissions and a test environment separate from production
  • Logs, monitoring, escalation and a continuity plan suited to the risk

Can the agents be customised?

Yes. Each agent is configured on the company's processes, approval rules, tone of voice and systems. Before building custom features we check inputs, outputs, integrations, permissions and exceptions; we do not build what is not needed.

  • Tone of voice: your brand's language and terminology.
  • Business processes: existing approval flows and decision rights.
  • Custom features: when the standard does not cover the use case.
  • Integrations: ERP, CRM, internal databases and existing software via APIs or connectors.

What kind of support is included?

Support is defined for each solution and the agreed service level. The quote spells out channels, hours, priorities, responsibilities and response times.

  • Agreed channels and availability window
  • Request classification and escalation path
  • An operational owner and review schedule
  • Monitoring, logs and the evidence needed for diagnosis
  • Training and documentation when included in the scope

How is data security checked?

Security is assessed on the actual scope of the solution and the vendors involved. Before go-live we document which data is processed, where it travels, who can access it and what evidence is available.

  • Data, purpose, location, retention and sub-processors
  • Encryption, identity, access control and separate environments
  • Logs, backups, incident response and business continuity
  • Deletion, export and exit clauses

Is there a trial period?

It depends on the solution. When useful, we set up a demo or a limited pilot to check the use case before extending the scope.

  • Sample or anonymised data consistent with the use case
  • Starting measurement and KPIs defined before the run
  • Normal cases, errors, exceptions and escalations included in the test
  • A documented final decision: extend, correct or stop

Is your website visible to AI assistants?

ChatGPT, Google AI Overviews and Perplexity answer with only a few sources. The free AI-readiness analysis checks how they read and cite your site and lists the priority fixes on content and structured data.

Check your AI visibility

Show us the process you want to automate

We assess together whether you need an agent, which one and with which checks. Sometimes the answer is that AI is not needed, and then we say so.

Show us a process

The first 30-minute call is free. We reply within 24 working hours.