How We Work

Challenge

A distribution company ran order data on one PostgreSQL server on aging access switches that dropped packets under load. Afternoon reporting slowed the order desk.

Approach

Replace the core and access switches, certify new Cat6A runs to the server row, and put storage and replication traffic on dedicated VLANs. Build a three-node PostgreSQL cluster with streaming replication and Patroni on RHEL hosts hardened to CIS Benchmarks, and move reporting to a read replica.

Design Targets

• Replication lag under 1 second
• Planned failover under 60 seconds
• Reporting isolated from order entry

Challenge

A professional-services firm moving floors needed hundreds of drops, a new server closet and six conferencing rooms live by Monday, with no collisions between electricians, cablers and AV installers.

Approach

One project plan across trades: power and pathways first, then cabling and rack build, then switch provisioning and AV commissioning. Certification reports for each link are delivered before move day, and the cutover is rehearsed the weekend before.

Design Targets

• Every link certified and labeled
• Rooms commissioned with documented control presets
• Cutover rehearsed before move weekend

Challenge

Three departments each reported revenue and resolved tickets differently, from separate databases and spreadsheets, and the numbers never matched in the Monday meeting.

Approach

Stage data from each system into a warehouse, model a star schema with conformed customer and date dimensions, and publish one governed Power BI semantic model with row-level security and scheduled incremental refresh.

Design Targets

• One definition for each metric
• Refresh inside the overnight window
• Regional leaders see only their own region

Challenge

A small internal IT team was stuck in reactive tickets, while Microsoft 365 permissions and backups had no clear owner.

Approach

ACG takes Tier 2 and 3 escalations and after-hours coverage, audits Microsoft 365 licensing and conditional access, deploys monitoring with alert routing, and adds immutable backups with scheduled restore tests.

Design Targets

• A named owner for every system
• Quarterly restore test on the calendar
• Patch calendar published to staff

Challenge

An acquirer wanted an independent read on a target company's IT before signing, covering the network, the perimeter and the system everyone complained about.

Approach

Review switch and firewall configurations against vendor hardening guidance, check perimeter exposure, and profile the busiest SQL Server workload for waits and missing indexes. Findings are ranked by risk and effort.

Design Targets

• Ranked findings with supporting evidence
• Remediation estimate by effort
• Executive summary in plain language

Challenge

A warehouse had a few analog cameras, keyed doors and no usable video after an after-hours break-in. Nobody could say who held a key.

Approach

Pull Cat6A to every camera and door, power the devices from managed Power-over-Ethernet switches on a dedicated VLAN, install IP cameras with on-site recording, and tie the badge readers to directory accounts so access ends the day someone leaves.

Design Targets

• Every door and camera documented
• Camera network isolated from office traffic
• Badge access removed the day an employee leaves

Challenge

A 60-person firm wanted AI help with support tickets and contracts but could not send client files to a public service.

Approach

Install a GPU server in the existing rack on an isolated network segment, run a local language model, index approved file shares with each person's own permissions, and route new tickets through a summarizer so a technician reviews every drafted reply.

Design Targets

• No client data leaves the network
• Answers respect each person's file permissions
• A technician approves every drafted reply