Solita

solita.fi

Solita combines technology advisory with implementation and operational support rather than operating solely as a strategy firm or software vendor. Its AI services cover discovery, technical planning, pilots, application development and lifecycle support. Buyers should distinguish these services from its proprietary FunctionAI platform and distinguish production references from proof-of-concept work. Its broader delivery scope includes custom software, cloud migrations, data platforms, APIs and cybersecurity assessments. ^[1] ^[2] ^[3] ^[4]

Services and delivery

AI discovery and technical planning

Advisory. Advises on AI opportunities, solution feasibility, implementation roadmaps and architecture. Planning can address algorithms, retrieval design, vector databases, LLMs, MLOps and orchestration. ^[1]

Data and AI strategy

Advisory. Connects business priorities with data and AI opportunities, capability assessments and development objectives. This is strategy advisory rather than a commitment to deploy software. ^[5]

AI pilots and embedded application development

Implementation. Builds AI pilots and embeds AI into applications and business processes. Production acceptance and rollout scope need to be defined for each engagement. ^[1]

Generative AI solution development

Implementation. Designs and develops GenAI proofs of concept, pilots and LLM solutions. The documented scope includes retrieval-augmented generation, chunking, embeddings, vector databases and orchestration. ^[2]

MLOps and LLMOps setup

Implementation. Establishes architecture and processes for managing AI application lifecycles. Buyers should specify evaluation, deployment and monitoring responsibilities separately from ongoing support. ^[2]

AI lifecycle operations

Managed services. Provides ongoing operation and lifecycle support for AI applications. Contract scope should distinguish model monitoring, application maintenance and underlying infrastructure management. ^[2]

Custom software development and modernisation

Implementation. Implements business-system upgrades, backends, critical registers and Android or iOS applications, with automated testing, deployment and continuous development. ^[6]

Cloud strategy and architecture advisory

Advisory. Advises on cloud strategy, governance, architecture and migration readiness. This planning work is distinct from migration execution. ^[7]

Cloud migration and engineering

Implementation. Executes cloud migrations, application modernisation and DevOps engineering. The service description names AWS, Microsoft Azure and Google Cloud. ^[7]

Managed cloud operations

Managed services. Manages cloud platforms and infrastructure through monitoring, incident handling, patching, backup and recovery, and cost optimisation. CloudBlox is presented as a managed-service concept. ^[8]

Data platform and data management implementation

Implementation. Implements data platforms, enterprise data warehouses, master-data management, catalogs and data-quality tools. This data engineering scope does not itself establish AI delivery. ^[9]

Integration and API implementation

Implementation. Designs and implements custom integrations, APIs, iPaaS setups and event-driven architectures, including testing. Post-go-live operations are also described on the service page. ^[10]

Cybersecurity assessment and advisory

Advisory. Provides threat assessments, secure-development reviews and penetration testing with remediation recommendations. Cloud-security architecture is also within the documented advisory scope. ^[11]

FunctionAI platform

Software product. A modular LLM platform for assistants, agents, workflows and applications, with orchestration, integrations, access management and deployment options. Platform procurement is distinct from associated advisory and implementation services. ^[4]

Company details

Technology and platforms

Named project technologies include Amazon Bedrock, Textract and Claude in Fimea’s reporting application, and Azure OpenAI Service in Tana’s support-assistant proof of concept. These references establish particular project stacks, not universal platform requirements. ^[12] ^[13]

Track record

Solita reports that Fimea’s retrieval-augmented application became a permanent part of its adverse-reaction reporting production environment. It retrieves drug information and produces referenced summaries for human reviewers rather than making decisions itself. ^[12]

For KONE, Solita describes joint development of cloud and edge analytics using elevator sensor data, including machine-learning and statistical fault-prediction models. The case reports global deployment but does not identify the model framework or cloud vendor. ^[14]

Tana’s cited technical-support assistant was explicitly a proof of concept using manuals and incident reports. Kela’s reference combines AI strategy, opportunity identification and small-scale experimentation; neither should be read as evidence of broad production deployment. ^[13] ^[15]

In the Tamro reference, Solita reports CloudBlox monitoring, 24/7/365 coverage, responsibility for updates and backups, and a service manager providing monthly reporting. This is a concrete managed-operations reference rather than cloud advice alone. ^[16]

Scope and limitations

The supplied evidence establishes commercial offerings and reported delivery, but not proposal-specific pricing, staffing, acceptance criteria or contractual support commitments.
Case-study outcomes are Solita’s reports, not independently verified measurements. Production deployments and proofs of concept should be evaluated separately.
FunctionAI licensing, proprietary dependencies and exit arrangements require verification. RoadCrewAO is documented as a delivery tool, but standalone procurement and licensing are unverified.
The precise division between AI model monitoring, application support and infrastructure operations is not established for a prospective engagement.

Questions to ask

Who owns architecture, implementation, production acceptance and ongoing model evaluation?
Which references match our use case and are production deployments rather than proofs of concept?
What retrieval-quality, hallucination and human-review acceptance thresholds will apply?
What data residency, model-provider and deployment constraints can the proposed architecture satisfy?
What support SLAs, retraining responsibilities, IP rights and exit arrangements will be contractual?
Are FunctionAI or other proprietary components optional, and what costs or dependencies would they introduce?

Overview