QwertyBit

QwertyBit is a Bucharest-founded AI agency built around a single idea: map your actual operations, find where time and cost leak, then ship custom agents to plug the holes. Founded in 2017 by Vlad Niculescu, a software engineer with a decade of production work in London's fintech and AI scenes, the firm operates as a small senior team working 1:1 with business owners. Vlad personally runs every engagement from discovery to production rollout. The agency has delivered 60+ projects since launch and works from offices in London and Bucharest.

Services and capabilities

QwertyBit offers four core services. AI agent development builds purpose-built agents for document processing, lead qualification, internal knowledge, compliance, and scheduling—shipped to production, not as demos. Business process automation maps workflows and automates repeatable steps with agents tied to measurable KPIs, covering sales ops, finance, compliance, and delivery pipelines. Custom software development delivers bespoke web, API, and mobile software for teams outgrowing off-the-shelf tools, with code landing in your GitHub from day one. LLM integration & deployment handles production deployments—cloud, on-prem, or hybrid—using Claude, GPT, Llama, Qwen, or Mistral, with eval harnesses, observability, and cost controls built in.

Every engagement follows the same four-phase methodology: Understand (map business holistically, surface time and cost leaks), Prioritise (score opportunities on impact, feasibility, risk; lock in transparent fixed-price quotes), Build (two-week release cadence; human approves sensitive actions; dashboards track cost, quality, errors), Optimise (monitor KPIs monthly, retrain on drift, retire agents that stop moving their target metric quarterly).

Engagements start with a free discovery meeting, then three total meetings over email and synchronously. A first working agent reaches production within 8–12 weeks. The agency ships in two-week increments, not big-bang deliveries.

Notable work

QwertyBit's portfolio spans regulated fintech, enterprise tech, B2B services, insurance, healthcare, logistics, real estate, legal, and analytics. A few examples with measured outcomes:

Fintech (KYC automation): A regulated financial services platform had manual KYC document review bottlenecking onboarding. QwertyBit built compliance-aware LLM workflows. Result: average onboarding time reduced by 64%, compliance analyst caseload up 3x without headcount increase.

Enterprise tech (internal knowledge): An engineering org faced constant interruptions from cross-department questions; answers lived scattered in Slack, Notion, Jira. Outcome: 51% faster onboarding for new developers, 3x faster internal support resolution time.

Data analytics & marketing (anomaly detection): Attribution data fragmented across five tools; campaign QA was manual and reactive (anomalies surfaced days later). Result: anomaly detection window from days to under 15 minutes, campaign QA time down 70%.

Insurance (contract risk): Manual contract review was slow and inconsistent. Outcome: 88% reduction in manual review time, risk-detection consistency 3x human baseline.

Sales automation (B2B services): Leads fell through cracks between qualification and follow-up. Result: 2.3x lead-to-deal conversion rate, 80% of follow-ups fully automated.

Real estate (document automation): Due-diligence required manual reading of hundreds of PDFs. Outcome: 90% reduction in manual document review, 3x faster project proposal turnaround.

Healthcare (scheduling): No-shows cost revenue; reminders were uniform, not targeted. Result: 34% reduction in no-show rates, 25% increase in daily appointment efficiency.

Logistics (invoice processing): Accounts payable spent hours on manual invoice keying. Outcome: 85% reduction in manual data entry, 60% faster invoice processing.

Legal (case summaries): Associates spent days on case review and memo drafting. Result: 60% time saved on document review, 3.5x faster client response times.

How they work

QwertyBit keeps engagement friction low. The Discovery meeting (60 minutes, free) walks through operations, data, tools, handoffs, and vendor spend to surface where time and cost leak. QwertyBit then sends back a ranked business case list by email. Prioritisation meeting reviews cases together; you agree what optimises first; transparent costs for each option go over email alongside scope, quote, and agreement. Kick-off meeting aligns on goals, milestones, tools, and cadence immediately after signing.

The Build phase ships two-week releases. A human approves anything risky; dashboards track cost per run, latency, error rate from day one. Data stays in your control; privacy is designed in, not bolted on. Code and models are yours.

Optimise phase monitors the KPI monthly, retrains if performance drifts, and brings new opportunities back to prioritisation quarterly. If an agent stops earning its keep, it is retired or reworked.

Pricing is transparent and fixed. You see build cost, projected monthly run cost, and the cost of doing nothing. No line items surprise you later.

Team and credentials

Vlad Niculescu founded QwertyBit in 2017 after shipping production systems across AI, fintech, and analytics in London. He holds an AI + Software Engineering degree from King's College London and personally runs every engagement. The firm is small and senior; exact headcount is not disclosed, but it operates as a tight team with Vlad at the helm. No third-party certifications (ISO, SOC 2, partner badges) are mentioned in the source material.