Services Digital practice Technology & AI

AI agents, automations and internal tools.

Working software, not slideware. We aim to have a first version running in your operations in about two weeks, on a fixed fee.

What we do

The work, plainly.

There's a market right now full of consultancies selling "AI strategy" to companies who really just need an inbox triage agent and a half-decent internal tool. We're the firm for the second group.

Our Digital practice exists because the Australian SMB middle market needs proper software work at sensible prices. Most weeks we're building one of three things: an LLM-backed agent that handles a high-volume manual workflow (inbox, supplier reconciliation, quoting), an internal portal that replaces three or four SaaS subscriptions, or a website that the client can actually edit themselves without ringing us.

We don't do enterprise transformation programs. We don't do "AI readiness assessments." We don't sell you a roadmap and walk away. The output is software that runs in your business on a date we wrote down in advance.

Capabilities

Five lanes of work.

Every Digital engagement falls into one of these five lanes, or some combination. If your need doesn't fit any of them, that's a signal we may not be the right firm and we'll say so on the call.

01

AI agents & RAG systems

Custom LLM agents reading inboxes, classifying tickets, drafting responses, routing exceptions. RAG pipelines on your internal docs, so every answer comes with its source.

02

Workflow automation

The boring middle layer: Xero ↔ CRM ↔ project tool ↔ payroll. We script the integrations that vendors won't, replace Zapier flows that have grown unwieldy, and remove the manual steps that nobody owns.

03

Internal tools & portals

A single internal app for a team currently rotating between four tabs and a paper folder. Built on Next.js or Rails depending on the team that'll maintain it — often replacing several overlapping SaaS subscriptions with one.

04

Websites & e-commerce

Marketing sites, lead funnels, Shopify and headless commerce. Built on stacks your future hires will recognise. We won't ship custom CMS work to a client who needs WordPress.

05

Cloud, data & IT support

AWS and Azure infra, data warehouse standup (BigQuery, Snowflake), backups that actually restore, and ongoing managed support if you don't want to hire an in-house team. Australian data residency by default.

How we work

Two-week sprints. One demo per Friday.

The engagement timeline is shared with every Valeron practice, but the Digital practice runs on a stricter cadence because software has more failure modes than a clean trial balance.

Sprint 0 · Day 1–5

Scope, write, sign.

We look at the system you're integrating with, write the actual sprint plan, and sign a fixed-fee SOW. No "discovery" billing.

Sprint 1 · Day 6–14

First demo, on the second Friday.

Working software in a staging environment by 4pm AEST. You see it, you push back on it, we ship to production by the following Tuesday unless something material needs rework.

Sprint n+ · Ongoing

Two-week cadence, indefinitely.

Each sprint ends with a deployed change. No "we'll ship next quarter" deferrals. If a sprint produces no shippable change, the partner explains why on a 15-minute call, in person.

Pricing

Two shapes. One fixed fee.

Single AI / software project: a fixed fee scoped to the work, quoted on the one-page proposal. Invoiced 50/50 at signing and delivery.

Ongoing Digital retainer: a flat monthly fee for continuous delivery plus a partner's oversight, sized to the scope you set on the discovery call.

Large enterprise transformation programs and enterprise SAP migrations aren't us. We'll refer you and stay friends.

Common questions

What clients ask before they sign on AI work.

Do you use OpenAI / Anthropic / open-source? How do you decide?
All three, plus open-source. We choose per engagement based on data sensitivity, latency, cost, and the long-term maintainability story. Anthropic for most agentic work; OpenAI for some structured-output cases; open-source where the data must not leave Australian infra. We document the choice in the SOW and revisit it at each quarterly review.
Where does our data go? Is it used to train models?
No model training on your data, ever. We use the no-training enterprise endpoints from OpenAI and Anthropic, or self-hosted open-source where required. All data resides in Australian regions of the cloud we deploy to (AWS Sydney by default).
What if the AI agent makes a mistake?
Every shipped agent has an explicit confidence threshold below which it asks a human. The agent's actions are logged with the inputs and the reasoning. We review the false-positive and false-negative rate at each quarterly review and tune. Hallucinations in customer-facing copy are caught by a second-model pass before send.
Can we own the code at the end?
Yes. Source is in your GitHub from day one. The IP belongs to you. If you part ways with us, you keep everything and we send a clean handover document.
Do you do RFP responses?
Rarely. RFPs reward firms that are good at writing RFP responses, which is not what we want to be good at. If you're running an RFP because procurement requires it, we'll respond briefly; if it's the substantive evaluation, a 45-minute call usually tells both of us more.

Book a partner-led call.

A Valeron technology partner will be on the call. Forty-five minutes. One-page proposal by Friday if it's a fit.