Ballast Group · AI consultancy for mid-sized businesses

We put AI to work inside your business, and we stay accountable for it.

Most companies don’t need more AI tools. They need someone to work out where AI genuinely pays for itself, build it properly, and take responsibility for keeping it safe once it’s running. That is the whole of what we do.

  • i. A private AI console, built around the way your business already works.
  • ii. A team that operates, monitors and improves it for you, every month.
  • iii. The option to own it outright, under your brand, whenever you’re ready.

Most AI projects stall for one of three reasons. We’re built around avoiding all three.

  1. 1

    No clear return

    Tools get bought because AI is urgent, not because anyone measured what it should change.

    We start with an assessment that names the work AI should take on, what it’s worth, and how we’ll know.

  2. 2

    Nobody trusts it

    Staff won’t rely on answers they can’t check, and leadership won’t sign off on risks nobody can explain.

    Every answer cites its source, every action is logged, and anything consequential waits for a person.

  3. 3

    Nobody owns it

    The pilot works, the consultant leaves, and six months later nothing has been updated or improved.

    We don’t hand over and walk away. We run it, month after month, and answer for how it performs.

01

The practice

One firm, responsible from the first conversation to the thousandth automated task.

Assessment
We study your workflows, data and risk exposure, then tell you plainly which uses of AI are worth doing, in what order, and which to leave alone.
Build
We design and engineer your console: assistants, automations and integrations with the systems you already use, shaped around your people rather than the other way round.
Safeguards
Access control, data protection, approvals and audit logging are part of the architecture from day one. Nothing goes live without them.
Operations
We host it, watch it, update it and keep improving it. Your team uses it; they never have to maintain it.

02

Where AI pays for itself

The work that is frequent, rule‑bound and text‑heavy, which describes a surprising amount of every business. For each, we’re specific about what stays with your people.

Customer support

Drafts accurate replies from your help centre, policies and past tickets, so agents review and send rather than write from scratch.

AI handles
Triage and tagging, first drafts, summaries of long threads, suggested help articles
People keep
Refunds, complaints, anything sensitive, and the final send
Connects to
Help desks such as Zendesk, Intercom or Freshdesk, and shared inboxes

Finance & back office

Reads invoices and receipts, matches them to orders, and flags what doesn’t add up, before anyone has to key it in.

AI handles
Document capture, matching, coding suggestions, anomaly flags, month‑end summaries
People keep
Every payment approval and every judgement call
Connects to
Accounting systems such as QuickBooks, Xero or NetSuite

Sales

Prepares your team before every conversation and keeps the CRM current afterwards, so selling time is spent selling.

AI handles
Account research, meeting briefs, call notes, CRM updates, first drafts of proposals
People keep
Pricing, commitments and every message that goes to a client
Connects to
CRMs such as HubSpot or Salesforce, plus email and calendar

Operations

Answers “how do we do this?” from your own procedures, and moves information between systems that don’t talk to each other.

AI handles
Procedure Q&A, recurring reports, data entry between systems, scheduling drafts
People keep
Exceptions, supplier decisions and anything safety‑critical
Connects to
Spreadsheets, shared drives, project tools and line‑of‑business systems

People & HR

Gives every employee instant, consistent answers about policies and benefits, and makes onboarding feel organised.

AI handles
Policy questions, onboarding guides, job description drafts, training material
People keep
Hiring, performance, pay, and any decision about an individual
Connects to
HR systems, handbooks and internal knowledge bases

03

The console

A single, private place where your company uses AI, governed by your rules and isolated from everyone else’s.

Your people

Staff, signed in through your identity provider
Customers, through channels you approve

Your console one isolated environment per client

Roles & permissions
Redaction of sensitive data
Human approval for high‑impact actions
Complete audit log
Your documents & systems, encrypted, never shared

AI models

Enterprise providers whose terms forbid training on your data
Chosen per task; replaceable without a rebuild
Every request passes through your console’s controls before it reaches a model, and every response is checked and logged on the way back.
Audit log Illustration
  1. AllowedSupport assistant drafted reply to ticket 4821 · sent by J. Alvarez
  2. RedactedCard number removed from customer email before model call
  3. HeldSupplier payment of $18,400 awaiting approval from Finance
  4. BlockedHidden instruction in uploaded PDF ignored and reported

What your team sees: assistants for support, sales, operations, finance and HR; a knowledge base that answers from your own documents and cites its sources; and automations that do the routine work and ask before doing anything that matters.

What your administrators see: who can use what, what it costs, what it has done, and a switch to pause any assistant immediately. It carries your name, your logo and your domain.

04

Safety standards

These are commitments, not features. Each one is written into your agreement with us.

  1. Your data is yours

    It is never used to train any model, ours or a provider’s. You can export or delete it at any time.

  2. Isolation by default

    Each client runs in a separate environment with its own encryption keys. No shared databases.

  3. Encryption throughout

    TLS in transit, AES‑256 at rest, and secrets held in a managed vault.

  4. Least privilege

    Single sign‑on, multi‑factor authentication, and each assistant limited to the data its job requires.

  5. Guarded inputs and outputs

    Sensitive data is redacted before it reaches a model; injected instructions in documents and emails are detected and refused.

  6. People decide

    Sending, paying, deleting and changing records require human approval. You set where that line sits.

  7. A complete record

    Every prompt, response and action is logged with who, what and when, and can be exported for audit.

  8. Tested before release

    Every change is evaluated against misuse, fabrication and data‑leak scenarios before your team sees it.

What we will never do

  • Sell, share or train on your data.
  • Let AI take an irreversible action you haven’t authorised.
  • Hold your data or configuration hostage.
  • Keep a security incident from you. If something happens, you hear it from us first.

Our data processing agreement and security documentation are available on request, under NDA if you prefer.

05

Ownership

You are never locked in. When you want more control, the console comes with you. We simply keep it running underneath.

Question Managed Co‑managed White‑label
Who builds new assistantsWe doYour administrators, with our supportYour team
Hosting, security & updatesUsUsUs, behind the scenes
Name on the consoleYoursYoursYours alone; no mention of us anywhere
Offer it to your own customers——Yes
Best suited toGetting startedTeams building in‑house skillCompanies ready to own their AI

Each arrangement is a fixed monthly fee, quoted after assessment. Moving between them is planned together, with no rebuild and no downtime.

06

How an engagement runs

Deliberately unhurried at the start, so nothing is rushed once it matters.

  1. First

    Conversation

    Thirty minutes on your business and what you want from it. No charge, and you’ll leave with useful ideas either way.

  2. Then

    Assessment

    A written roadmap: the opportunities we found, the risks we see, and what we recommend you do first.

  3. Next

    Pilot

    One high‑value use case, live in your console with a small group, with every safeguard in place.

  4. After

    Rollout

    We extend it across teams and train your people to use it well and use it safely.

  5. Monthly

    Operation

    Monitoring, improvements, new use cases and a regular safety review, reported to you in plain language.

We started this firm because we kept seeing the same thing: capable businesses either avoiding AI out of caution, or adopting it with no one minding the risks. Neither is necessary. AI can be put to work carefully, and someone should be answerable when it is. We named the firm Ballast because ballast is the weight that keeps something steady under pressure. That’s the job we’ve chosen.

The founders, Ballast

07

Questions

Do we need technical staff?

No. We build, host and run everything. If you later want your own people building systems, we’ll train them.

What if the AI gets something wrong?

Consoles are designed so that AI proposes and people decide on anything consequential. Answers cite their sources, important actions wait for approval, everything is logged, and any system can be paused at once.

What happens if we want to stop working with you?

You can move to a white‑label arrangement and keep running the console under your own name, or export your data and configuration and leave. Either way, it belongs to you.

Which AI models do you use?

We choose among leading enterprise providers for each task, weighing quality, cost and your compliance requirements, and can change provider without you rebuilding anything.

Is our industry too regulated for this?

Regulated businesses are often where careful AI helps most, because so much of the work is documentation. We design around your obligations from the assessment onward, and will tell you plainly if something shouldn’t be automated.

How is it priced?

Assessments start at $3,500 and are credited against the build if you go ahead within 60 days. Managed plans start at $1,500 a month, with AI usage included up to a cap you set, so there are no surprise charges. Every plan runs for at least 12 months from launch day, because results build month by month.

Who do you work with?

US businesses with roughly 50 to 500 people, especially those with a lot of client communication and document work: accounting and law firms, insurance and benefits agencies, property managers, contractors and distributors. The systems are built around how your business already works, so the industry matters less than the shape of the work.

How long until it’s running?

The assessment takes one to three weeks. After you sign, the build is planned to go live in about seven weeks: a week of setup and access, then building and testing, a one-week pilot with a small group, and training in the launch week.

Is our data used to train AI models?

No. Each client gets a private, isolated environment, we use enterprise AI services that do not train on customer data, and we connect only to the tools a task needs. You can export your data at any time, and if you leave we delete your environment and confirm it in writing.

Start with a conversation.

Tell us a little about your business and we’ll arrange a thirty‑minute call. If you’d rather sign an NDA before discussing details, just say so.

  • Reply within one business day
  • No charge for the first conversation
  • No obligation, and no sales sequence afterwards

Or write to us directly at hello@ballastgroup.ai

Please don’t include passwords or confidential customer information.