Sales
Lead follow-up and qualification
Every new enquiry gets a reply by text or email within minutes, a few qualifying questions, and a booked call if it is a good fit.
Voice/chat agent
Curious Context Labs
AI is powerful.
It doesn't know your business.
Most AI projects stall because the AI is missing context, not because the model is weak. We learn how your business works, then design, build and run the AI that does the work with you.
30 minutes, free. You talk, we ask questions. No slides.
chapter 00 the gap
The model is not the problem. What's missing is context.
A generic chatbot can write a polite reply to your customer. It does not know:
- your refund ruleunknown
- how much stock you haveunknown
- that this customer is your biggest accountunknown
That is where most AI projects stall. Not because the model is weak, but because nobody gave it the context.
chapter 01 curious
Curious.We start with questions, not tools.
Before we suggest any AI, we talk to the people who do the work. We ask what they check, who they wait for and where things go wrong. Then we map how the work really flows between people and tools.
Questions we ask in every interview
- Where does your week actually go?
- What do you check before you reply?
- Who has to say yes, and how long does that take?
- Where do mistakes creep in?
- What would you stop doing tomorrow if you could?
chapter 02 context
Context.We give the AI what it's missing.
- What it knows
- How it decides
- What's happening now
An AI model knows a lot about the world and nothing about your business. We connect it to three kinds of context, and only the parts each task needs.
- What the business knows
- SOPs, policies, pricing rules and past work.
- How it decides
- Approvals, exceptions and limits.
- What is happening right now
- Orders, tickets, pipeline and the inbox.
One question, three kinds of context
A customer asks: “Can I return this?”
To answer well, the AI needs:
- KnowsThe returns policy and its time limit
- DecidesThe exceptions, and who can approve one
- NowThis order's delivery date, from your store system
With all three, the AI gives the right answer or hands the case to a person. With only the model, it guesses.
chapter 03 labs
Labs.We prototype fast, test on real cases and ship.
- Command: run invoice-intake --source accounts-inbox
- reading the accounts inbox: 14 new emails
- finding invoices in attachments: 6 invoices
- extracting supplier, amount, due date: 6 of 6
- matching to purchase orders: 5 matched
- mismatch on INV-2291: billed for 120 units, PO-0874 ordered 100. Held for review.
- drafting bills in Xero (not posted): 5 drafts
- waiting for approval from the finance lead
- approved by finance lead: 5 bills posted, 1 held with a note
A sample agent run with illustrative data. Nothing is posted until a person says yes.
We work like a lab: small experiments, real data, honest results. You see the system working early, and it only goes live once it has passed tests on your own past cases.
How we work in the lab
An early demo on your own data
Not a slide deck. You see it handle your real documents and cases while there is still time to change course.
Tested on real past cases
We run it on work your team has already done, where the right answer is known, and fix what it gets wrong.
A person approves what matters
Payments, customer promises, anything legal: the AI prepares the work, a person decides.
Wrong answers become fixes
When it gets a case wrong, we find out why, add the missing context and run the tests again.
chapter 04 what changes
What changes.
Here is what we aim for, in plain words. Real numbers from client projects will sit next to each one.
Time back
Hours of copying, chasing and summarising move to the AI. People spend the time on customers and judgement calls.
[CASE STUDY]Real numbers from a client project go here.
Faster replies
Customers, leads and staff get answers in minutes, including at night and on weekends.
[CASE STUDY]Real numbers from a client project go here.
Fewer errors
The AI checks every invoice, order and contract against your rules, and flags what does not match.
[CASE STUDY]Real numbers from a client project go here.
What we do
One team, from the first question to a system that runs.
The people who work out where AI fits are the same people who build it and keep it running. Use one service, or all three in order.
01 advise
Advise
Find where AI pays off.
We interview the people who do the work, map how it really flows, and rank the places where AI would save time or money. You get a plan you can act on, with or without us.
- Interviews with the people who do the work
- A map of your key workflows and the systems behind them
- Ranked use cases with effort, risk and expected benefit in plain words
- A data and security check: what the AI can see, and what it must not
- A build plan with fixed prices for the first projects
deliverable: AI audit and roadmap
02 build
Build
Ship working AI, at a fixed price.
We design and build the agents, assistants, apps and automations from your roadmap. You see an early demo on your own data, and we agree the price before we start.
- AI agents that work across your tools and ask a person to approve
- A company AI assistant that answers from your own documents
- Custom AI apps for one job, built for the people who do it
- Voice and chat agents for customers and staff
- Automations between the tools you already pay for
deliverable: A working system, handed over with code, prompts and docs
03 run
Run
Keep it working as you grow.
AI systems need care: models change, data drifts, teams find new uses. We monitor what we built, fix issues, update models, train your team and add new use cases each month.
- Monitoring of accuracy, cost and usage
- Fixes and model updates when providers change
- Training sessions and short guides for your team
- New use cases added from your backlog
- A monthly note on what the AI did and what to do next
deliverable: AI partner retainer
ls ./what-we-build
- Agent
- Company assistant
- Custom app
- Voice/chat agent
- Automation
Use cases
Pick a department. See what AI can take off its plate.
Each use case is written up step by step: what the AI does, where a person approves, and which tools it connects to.
73 use cases across 10 departments
sales: 7 use cases
Follow up faster, write proposals in minutes and keep the CRM honest.
Sales
Proposal and quote drafting
Turns call notes and site details into a priced first-draft proposal using your own templates, rate card and past wins, ready for a salesperson to check.
AgentSales
Pre-meeting account briefs
Before each sales call, reps get a one-page brief on the account: who is attending, past emails, open deals, support issues and recent company news.
Company assistant
Integrations
Works with the tools you already use
We connect AI to the systems your team works in every day, so nobody has to learn a new app or copy data between screens.
How we work
Four steps. You always know what happens next.
step 01, 30 minutes, free
Talk
You tell us how the business runs. We ask a lot of questions.
- We ask where time goes, where errors creep in and where customers wait.
- You leave with a few ideas, and an honest view on whether AI is the right fix.
step 02, 1–2 weeks
Map
We map the work and pick the use cases that pay off.
- We interview the people who do the work and trace how it flows between tools.
- You get a ranked roadmap and a fixed price for each build.
step 03, Fixed price
Build
We build, show you an early demo on your data, and ship.
- You see it working on real cases early, so there are no surprises at the end.
- A person approves anything that matters before the AI acts on it.
step 04, Monthly
Run
We monitor, improve and add new use cases.
- We watch quality and cost, update models and fix issues.
- We train your team and keep a backlog of what to automate next.
Why us
Most AI projects fail on context. We start there.
We start with questions
Most AI projects fail on context, not technology. We spend the first days learning how your business actually works.
Builders, not slide-makers
The people who advise you are the people who build the system. Nothing gets lost between strategy and code.
Work moves outside your office hours
Our team is in India. We keep fixed overlap hours for calls with the US, UK and Australia, and much of the build work happens while your office is closed.
Small-business sized
We work with small and mid-sized teams. Projects are scoped to pay back, not to fill a budget.
protocol.txtrev 1 · read-only
Four promises we keep on every project
Vendor-independent
OpenAI, Claude, Gemini or open models — we choose per task, based on quality, cost and where your data may go.
You own everything
The code, prompts, data and AI accounts are yours. If we part ways, nothing is locked in with us.
A person approves what matters
Payments, customer promises, hiring decisions, anything legal: the AI drafts, a person decides.
Fixed price before we build
We agree the scope and price up front. If we underestimated, that is our problem, not yours.
signed: [NAMES AND ROLES]
for: Curious Context Labs
Engagements and pricing
Start free. Know the price before we build.
Every audit and build has a fixed price agreed up front. The retainer is monthly and you can stop it with 30 days' notice.
Free consultation
time: 30 minutes
price: Free
Tell us how your business runs and where work piles up. We ask questions and tell you honestly whether AI can help.
Book a free consultationAI audit and roadmap
time: 1–2 weeks
price: From [PRICE]
We map your workflows, find the use cases that pay off and price the first builds.
Book a call to scope an auditBuild project
time: Usually 3–8 weeks
price: From [PRICE]
Fixed price, agreed before we start
We build one system end to end, show you an early demo on your own data, and hand it over.
Book a call to discuss a buildAI partner retainer
time: Monthly
price: From [PRICE]/month
Cancel with 30 days' notice
We keep your AI working, train your team and add new use cases every month.
Book a call about a retainer
Questions
What business owners ask us first
Something else on your mind? Book a free consultation and ask us directly.
We are a small business. Is AI worth it for us?
Often, yes — small teams feel repetitive work the most. The free consultation is there to find out. If AI is not the right fix, we will say so.
Which AI models do you use?
Whichever fits the task: OpenAI, Claude, Gemini or open models. We compare quality, cost and data rules for each job, and we can switch later without rebuilding everything.
Is our data safe?
We give the AI only the data a task needs, use business-grade AI accounts that do not train on your data, and keep those accounts in your name. We document what the system can see and do.
Who owns what you build?
You do. The code, prompts, data and AI accounts belong to you from day one.
Will the AI act on its own?
Only where you are comfortable. For anything that matters — money, customer commitments, hiring, legal — the AI prepares the work and a person approves it.
Your team is in India. How do calls work?
We keep fixed overlap hours with US, UK and Australian time zones for calls and reviews. Much of the build work happens outside your office hours, so there is usually progress to review when you next check in.
How long does a build take?
Most first builds take three to eight weeks. You see an early demo on your own data well before the end.
What does it cost?
The consultation is free. Audits, builds and retainers have fixed prices agreed before we start. See the pricing page for starting points.
What tools can you connect to?
Most tools with an API or export: CRMs like HubSpot and Salesforce, accounting tools like Xero and QuickBooks, Google Workspace, Microsoft 365, Slack, Teams, helpdesks and industry systems.
Tell us how your business runs. We'll tell you where AI fits.
A free 30-minute call. You talk, we ask questions, and you leave with a few concrete ideas — or an honest answer that AI isn't the right fix yet.
Working with teams in the United States, United Kingdom and Australia.